Dissertations / Theses on the topic 'Neighbor selection'

It is important to detect faults in bus details at an early stage. Because the driving style affects the breakdown of different details in the bus, identification of the driving style is important to minimize the number of failures in buses.

The identification of the driving style of the driver was based on the input data which contained examples of the driving runs of each class. K-nearest neighbor and neural networks algorithms were used. Different models were tested.

It was shown that the results depend on the selected driving runs. A hypothesis was suggested that the examples from different driving runs have different parameters which affect the results of the classification.

The best results were achieved by using a subset of variables chosen with help of the forward feature selection procedure. The percent of correct classifications is about 89-90 % for the k-nearest neighbor algorithm and 88-93 % for the neural networks.

Feature selection allowed a significant improvement in the results of the k-nearest neighbor algorithm and in the results of the neural networks algorithm received for the case when the training and testing data sets were selected from the different driving runs. On the other hand, feature selection did not affect the results received with the neural networks for the case when the training and testing data sets were selected from the same driving runs.

Another way to improve the results is to use smoothing. Computing the average class among a number of consequent examples allowed achieving a decrease in the error.

Thesis advisor: Arthur Lewbel
This dissertation consists of three independent essays in applied microeconomics and econometrics. Essay 1 investigates the issue why individuals with health insurance use more health care. One obvious reason is that health care is cheaper for the insured. But additionally, having insurance can encourage unhealthy behavior via moral hazard. The effect of health insurance on medical utilization has been extensively studied; however, previous work has mostly ignored the effect of insurance on behavior and how that in turn affects medical utilization. This essay examines these distinct effects. The increased medical utilization due to reduced prices may help the insured maintain good health, while that due to increased unhealthy behavior does not, so distinguishing these two effects has important policy implications. A two-period dynamic forward-looking model is constructed to derive the structural causal relationships among the decision to buy insurance, health behaviors (drinking, smoking, and exercise), and medical utilization. The model shows how exogenous changes in insurance prices and past behaviors can identify the direct and indirect effects of insurance on medical utilization. An empirical analysis also distinguishes between intensive and extensive margins (e.g., changes in the number of drinkers vs. the amount of alcohol consumed) of the insurance effect, which turns out to be empirically important. Health insurance is found to encourage less healthy behavior, particularly heavy drinking, but this does not yield a short term perceptible increase in doctor or hospital visits. The effects of health insurance are primarily found at the intensive margin, e.g., health insurance may not cause a non-drinker to take up drinking, while it encourages a heavy drinker to drink even more. These results suggest that to counteract behavioral moral hazard, health insurance should be coupled with incentives that target individuals who currently engage in unhealthy behaviors, such as heavy drinkers. Essay 2 examines the effect of repeating kindergarten on the retained children's academic performance. Although most existing research concludes that grade retention generates no benefits for retainees' later academic performance, holding low achieving children back has been a popular practice for decades. Drawing on a recently collected nationally representative data set in the US, this paper estimates the causal effect of kindergarten retention on the retained children's later academic performance. Since children are observed being held back only when they enroll in schools that permit retention, this paper jointly models 1) the decision of entering a school allowing for kindergarten retention, 2) the decision of undergoing a retention treatment in kindergarten, and 3) children's academic performance in higher grades. The retention treatment is modeled as a binary choice with sample selection. The outcome equations are linear regressions including the kindergarten retention dummy as an endogenous regressor with a correlated random coefficient. A control function estimator is developed for estimating the resulting double-hurdle treatment model, which allows for unobserved heterogeneity in the retention effect. As a comparison, a nonparametric bias-corrected nearest neighbor matching estimator is also implemented. Holding children back in kindergarten is found to have positive but diminishing effects on their academic performance up to the third grade. Essay 3 proves the semiparametric identification of a binary choice model having an endogenous regressor without relying on outside instruments. A simple estimator and a test for endogeneity are provided based on this identification. These results are applied to analyze working age male's migration within the US, where labor income is potentially endogenous. Identification relies on the fact that the migration probability among workers is close to linear in age while labor income is nonlinear in age(when both are nonparametrically estimated). Using data from the PSID, this study finds that labor income is endogenous and that ignoring this endogeneity leads to downward bias in the estimated effect of labor income on the migration probability
Thesis (PhD) — Boston College, 2009
Submitted to: Boston College. Graduate School of Arts and Sciences
Discipline: Economics

Data that is represented with high dimensionality presents a computational complexity challenge for many existing algorithms. Limiting dimensionality by discarding attributes is sometimes a poor solution to this problem because significant high-level concepts may be encoded in the data across many or all of the attributes. Non-linear dimensionality reduction (NLDR) techniques have been successful with many problems at minimizing dimensionality while preserving intrinsic high-level concepts that are encoded with varying combinations of attributes. Unfortunately, many challenges remain with existing NLDR techniques, including excessive computational requirements, an inability to benefit from prior knowledge, and an inability to handle certain difficult conditions that occur in data with many real-world problems. Further, certain practical factors have limited advancement in NLDR, such as a lack of clarity regarding suitable applications for NLDR, and a general inavailability of efficient implementations of complex algorithms. This dissertation presents a collection of papers that advance the state of NLDR in each of these areas. Contributions of this dissertation include: • An NLDR algorithm, called Manifold Sculpting, that optimizes its solution using graduated optimization. This approach enables it to obtain better results than methods that only optimize an approximate problem. Additionally, Manifold Sculpting can benefit from prior knowledge about the problem. • An intelligent neighbor-finding technique called SAFFRON that improves the breadth of problems that existing NLDR techniques can handle. • A neighborhood refinement technique called CycleCut that further increases the robustness of existing NLDR techniques, and that can work in conjunction with SAFFRON to solve difficult problems. • Demonstrations of specific applications for NLDR techniques, including the estimation of state within dynamical systems, training of recurrent neural networks, and imputing missing values in data. • An open source toolkit containing each of the techniques described in this dissertation, as well as several existing NLDR algorithms, and other useful machine learning methods.

A presente dissertação propõe métodos para mineração de dados para diagnóstico de câncer de mama (CM) baseado na seleção de variáveis. Partindo-se de uma revisão sistemática, sugere-se um método para a seleção de variáveis para classificação das observações (pacientes) em duas classes de resultado, benigno ou maligno, baseado na análise citopatológica de amostras de célula da mama de pacientes. O método de seleção de variáveis para categorização das observações baseia-se em 4 passos operacionais: (i) dividir o banco de dados original em porções de treino e de teste, e aplicar a ACP (Análise de Componentes Principais) na porção de treino; (ii) gerar índices de importância das variáveis baseados nos pesos da ACP e na percentagem da variância explicada pelos componentes retidos; (iii) classificar a porção de treino utilizando as técnicas KVP (k-vizinhos mais próximos) ou AD (Análise Discriminante). Em seguida eliminar a variável com o menor índice de importância, classificar o banco de dados novamente e calcular a acurácia de classificação; continuar tal processo iterativo até restar uma variável; e (iv) selecionar o subgrupo de variáveis responsável pela máxima acurácia de classificação e classificar a porção de teste utilizando tais variáveis. Quando aplicado ao WBCD (Wisconsin Breast Cancer Database), o método proposto apresentou acurácia média de 97,77%, retendo uma média de 5,8 variáveis. Uma variação do método é proposta, utilizando quatro diferentes tipos de kernels polinomiais para remapear o banco de dados original; os passos (i) a (iv) acima descritos são então aplicados aos kernels propostos. Ao aplicar-se a variação do método ao WBCD, obteve-se acurácia média de 98,09%, retendo uma média de 17,24 variáveis de um total de 54 variáveis geradas pelo kernel polinomial recomendado. O método proposto pode auxiliar o médico na elaboração do diagnóstico, selecionando um menor número de variáveis (envolvidas na tomada de decisão) com a maior acurácia, obtendo assim o maior acerto possível.
This dissertation presents a data mining method for breast cancer (BC) diagnosis based on selected features. We first carried out a systematic literature review, and then suggested a method for feature selection and classification of observations, i.e., patients, into benign or malignant classes based on patients’ breast tissue measures. The proposed method relies on four operational steps: (i) split the original dataset into training and testing sets and apply PCA (Principal Component Analysis) on the training set; (ii) generate attribute importance indices based on PCA weights and percent of variance explained by the retained components; (iii) classify the training set using KNN (k-Nearest Neighbor) or DA (Discriminant Analysis) techniques, eliminate irrelevant features and compute the classification accuracy. Next, eliminate the feature with the lowest importance index, classify the dataset, and re-compute the accuracy. Continue such iterative process until one feature is left; and (iv) choose the subset of features yielding the maximum classification accuracy, and classify the testing set based on those features. When applied to the WBCD (Wisconsin Breast Cancer Database), the proposed method led to average 97.77% accurate classifications while retaining average 5.8 features. One variation of the proposed method is presented based on four different types of polynomial kernels aimed at remapping the original database; steps (i) to (iv) are then applied to such kernels. When applied to the WBCD, the proposed modification increased average accuracy to 98.09% while retaining average of 17.24 features from the 54 variables generated by the recommended kernel. The proposed method can assist the physician in making the diagnosis, selecting a smaller number of variables (involved in the decision-making) with greater accuracy, thereby obtaining the highest possible accuracy.

A análise de dados contendo informações sequenciais é um problema de crescente interesse devido à grande quantidade de informação que é gerada, entre outros, em processos de monitoramento. As séries temporais são um dos tipos mais comuns de dados sequenciais e consistem em observações ao longo do tempo. O algoritmo k-Nearest Neighbor - Time Series Prediction kNN-TSP é um método de previsão de dados temporais. A principal vantagem do algoritmo é a sua simplicidade, e a sua aplicabilidade na análise de séries temporais não-lineares e na previsão de comportamentos sazonais. Entretanto, ainda que ele frequentemente encontre as melhores previsões para séries temporais parcialmente periódicas, várias questões relacionadas com a determinação de seus parâmetros continuam em aberto. Este trabalho, foca-se em dois desses parâmetros, relacionados com a seleção de vizinhos mais próximos e a função de previsão. Para isso, é proposta uma abordagem simples para selecionar vizinhos mais próximos que considera a similaridade e a distância temporal de modo a selecionar os padrões mais similares e mais recentes. Também é proposta uma função de previsão que tem a propriedade de manter bom desempenho na presença de padrões em níveis diferentes da série temporal. Esses parâmetros foram avaliados empiricamente utilizando várias séries temporais, inclusive caóticas, bem como séries temporais reais referentes a variáveis ambientais do reservatório de Itaipu, disponibilizadas pela Itaipu Binacional. Três variáveis limnológicas fortemente correlacionadas são consideradas nos experimentos de previsão: temperatura da água, temperatura do ar e oxigênio dissolvido. Uma análise de correlação é realizada para verificar se os dados previstos mantem a correlação das variáveis. Os resultados mostram que, o critério de seleção de vizinhos próximos e a função de previsão, propostos neste trabalho, são promissores
Treating data that contains sequential information is an important problem that arises during the data mining process. Time series constitute a popular class of sequential data, where records are indexed by time. The k-Nearest Neighbor - Time Series Prediction kNN-TSP method is an approximator for time series prediction problems. The main advantage of this approximator is its simplicity, and is often used in nonlinear time series analysis for prediction of seasonal time series. Although kNN-TSP often finds the best fit for nearly periodic time series forecasting, some problems related to how to determine its parameters still remain. In this work, we focus in two of these parameters: the determination of the nearest neighbours and the prediction function. To this end, we propose a simple approach to select the nearest neighbours, where time is indirectly taken into account by the similarity measure, and a prediction function which is not disturbed in the presence of patterns at different levels of the time series. Both parameters were empirically evaluated on several artificial time series, including chaotic time series, as well as on a real time series related to several environmental variables from the Itaipu reservoir, made available by Itaipu Binacional. Three of the most correlated limnological variables were considered in the experiments carried out on the real time series: water temperature, air temperature and dissolved oxygen. Analyses of correlation were also accomplished to verify if the predicted variables values maintain similar correlation as the original ones. Results show that both proposals, the one related to the determination of the nearest neighbours as well as the one related to the prediction function, are promising

Today many car manufactures are developing systems that help the driver to avoid collisions. Examples of this kind of systems are: adaptive cruise control, collision warning and collision mitigation / avoidance.

All these systems need to track and predict future positions of surrounding objects (vehicles ahead of the system host vehicle), to calculate the risk of a future collision. To validate that a prediction is correct the predictions must be correlated to observations. This is called the data association problem. If a prediction can be correlated to an observation, this observation is used for updating the tracking filter. This process maintains the low uncertainty level for the track.

From the work behind this thesis, it has been found that a sequential nearest- neighbour approach for the solution of the problem to correlate an observation to a prediction can be used to find the solution to the data association problem.

Since the computational power for the collision mitigation system is limited, only the most dangerous surrounding objects can be tracked and predicted. Therefore, an algorithm that classifies and selects the most critical measurements is developed. The classification into order of potential risk can be done using the measurements that come from an observed object.

The present thrust in the electronics industry towards integrating multiple functions on a single chip while operating at very high frequencies has highlighted the need for efficient Electronic Design Automation (EDA) tools to shorten the design cycle and capture market windows. However, the increasing complexity in modern circuit design has made simulation a computationally cumbersome task. The notion of model order reduction has emerged as an effective tool to address this difficulty. Typically, there are numerous approaches and several issues involved in the implementation of model-order reduction techniques. Among the important ones of those issues is the problem of determining a suitable order (or size) for the reduced system. An optimal order would be the minimal order that enables the reduced system to capture the behavior of the original (more complex and larger) system up to a user-defined frequency. The contribution presented in this thesis describes a new approach aimed at determining the order of the reduced system. The proposed approach is based on approximating the impulse response of the original system in the time-domain. The core methodology in obtaining that approximation is based on numerically inverting the Laplace-domain of the representation of the impulse response from the complex-domain (s-domain) into the time-domain. The main advantage of the proposed approach is that it allows the order selection algorithm to operate directly on the time-domain form of the impulse response. It is well-known that numerically generating the impulse response in the time-domain is very difficult and its not impossible, since it requires driving the original network with the Dirac-delta function, which is a mathematical abstraction rather than a concrete waveform that can be implemented on a digital computer. However, such a difficulty is avoided in the proposed approach since it uses the Laplace-domain image of the impulse response to obtain its time-domain representation. The numerical simulations presented in the thesis demonstrate that using the time-domain waveform of the impulse response, computed using the proposed approach and properly filtered with a Butterworth filter, guides the order selection algorithm to select a smaller order, i.e., the reduced system becomes more compact in size. The phrase "smaller or more compact" in this context refers to the comparison with existing techniques currently in use, which seek to generate some form of time-domain approximations for the impulse response through driving the original network with pulse-shaped function (e.g., Gaussian pulse).

“Machine learning is the process of discovering and interpreting meaningful information, such as new correlations, patterns and trends by sifting through large amounts of data stored in repositories, using pattern recognition technologies as well as statistical and mathematical techniques” (Larose, 2005). From my understanding, machine learning is a process of using different analysis techniques to observe previously unknown, potentially meaningful information, and discover strong patterns and relationships from a large dataset. Professor Kasabov (2007b) classified computational models into three categories (e.g. global, local, and personalised) which have been widespread and used in the areas of data analysis and decision support in general, and in the areas of medicine and bioinformatics in particular. Most recently, the concept of personalised modelling has been widely applied to various disciplines such as personalised medicine, personalised drug design for known diseases (e.g. cancer, diabetes, brain disease, etc.) as well as for other modelling problems in ecology, business, finance, crime prevention, and so on. The philosophy behind the personalised modelling approach is that every person is different from others, thus he/she will benefit from having a personalised model and treatment. However, personalised modelling is not without issues, such as defining the correct number of neighbours or defining an appropriate number of features. As a result, the principal goal of this research is to study and address these issues and to create a novel framework and system for personalised modelling. The framework would allow users to select and optimise the most important features and nearest neighbours for a new input sample in relation to a certain problem based on a weighted variable distance measure in order to obtain more precise prognostic accuracy and personalised knowledge, when compared with global modelling and local modelling approaches.

SILVA, C. F. M. Contemporary electromagnetic spectrum reuse techniques: tv white spaces and D2D communications. 128 f. 2015. Tese (Doutorado em Engenharia de Teleinformática) – Centro de Tecnologia, Universidade Federal do Ceará, Fortaleza, 2015.
Submitted by Marlene Sousa (mmarlene@ufc.br) on 2016-03-29T12:13:57Z No. of bitstreams: 1 2015_tese_cfmsilva.pdf: 8667655 bytes, checksum: a7e392b18bd01de3be64a22b8ff9906b (MD5)
Approved for entry into archive by Marlene Sousa(mmarlene@ufc.br) on 2016-03-31T14:31:46Z (GMT) No. of bitstreams: 1 2015_tese_cfmsilva.pdf: 8667655 bytes, checksum: a7e392b18bd01de3be64a22b8ff9906b (MD5)
Made available in DSpace on 2016-03-31T14:31:46Z (GMT). No. of bitstreams: 1 2015_tese_cfmsilva.pdf: 8667655 bytes, checksum: a7e392b18bd01de3be64a22b8ff9906b (MD5) Previous issue date: 2015-12-12
Over the last years, the wireless broadband access has achieved a tremendous success. With that, the telecommunications industry has faced very important changes in terms of technology, heterogeneity, kind of applications, and massive usage (virtual data tsunami) derived from the introduction of smartphones and tablets; or even in terms of market structure and its main players/actors. Nonetheless, it is well-known that the electromagnetic spectrum is a scarce resource, being already fully occupied (or at least reserved for certain applications). Tra- ditional spectrum markets (where big monopolies dominate) and static spectrum management originated a paradoxal situation: the spectrum is occupied without actually being used! In one hand, with the global transition from analog to digital Television (TV), part of the spectrum previously licensed for TV is freed and geographically interleaved, originating the consequent Television White Spaces (TVWS); on the other hand, the direct communications between devices, commonly referred as Device-to-Device (D2D) communications, are attracting crescent attention by the scientific community and industry in order to overcome the scarcity problem and satisfy the increasing demand for extra capacity. As such, this thesis is divided in two main parts: (a) Spectrum market for TVWS: where a SWOT analysis for the use of TVWS is performed giving some highlights in the directions/actions that shall be followed so that its adoption becomes effective; and a tecno-economic evaluation study is done considering as a use-case a typical European city, showing the potential money savings that operators may reach if they adopt by the use of TVWS in a flexible market manner; (b) D2D communications: where a neighbor discovery technique for D2D communications is proposed in the single-cell scenario and further extended for the multi-cell case; and an interference mitigation algorithm based on the intelligent selection of Downlink (DL) or Uplink (UL) band for D2D communications underlaying cellular networks. A summary of the principal conclusions is as follows: (a) The TVWS defenders shall focus on the promotion of a real-time secondary spectrum market, where through the correct implementation of policies for protection ratios in the spectrum broker and geo-location database, incumbents are protected against interference; (b) It became evident that an operator would recover its investment around one year earlier if it chooses to deploy the network following a flexible spectrum market approach with an additional TVWS carrier, instead of the traditional market; (c) With the proposed neighbor discovery technique the time to detect all neighbors per Mobile Station (MS) is significantly reduced, letting more time for the actual data transmission; and the power of MS consumed during the discovery process is also reduced because the main processing is done at the Base Station (BS), while the MS needs to ensure that D2D communication is possible just before the session establishment; (d) Despite being a simple concept, band selection improves the gains of cellular communications and limits the gains of D2D communications, regardless the position within the cell where D2D communications happen, providing a trade-off between system performance and interference mitigation.
Nos últimos anos, o acesso de banda larga atingiu um grande sucesso. Com isso, a indústria das telecomunicações passou por importantes transformações em termos de tecnologia, heterogeneidade, tipo de aplicações e uso massivo (tsunami virtual de dados) em consequência da introdução dos smartphones e tablets; ou até mesmo na estrutura de mercado e os seus principais jogadores/atores. Porém, é sabido que o espectro electromagnético é um recurso limitado, estando já ocupado (ou pelo menos reservado para alguma aplicação). O mercado tradicional de espectro (onde os grandes monopólios dominam) e o seu gerenciamento estático contribuíram para essa situação paradoxal: o espectro está ocupado mas não está sendo usado! Por um lado, com a transição mundial da Televisão (TV) analógica para a digital, parte do espectro anteriormente licenciado para a TV é libertado e geograficamente multiplexado para evitar a interferência entre sinais de torres vizinhas, dando origem a «espaços em branco» na frequência da TV ou Television White Spaces (TVWS); por outro lado, as comunicações diretas entre usuários, designadas por comunicações diretas Dispositivo-a-Dispositivo (D2D), está gerando um crescente interesse da comunidade científica e indústria, com vista a ultrapassar o problema da escassez de espectro e satisfazer a crescente demanda por capacidade extra. Assim, a tese está dividida em duas partes principais: (a) Mercado de espectro eletromagnético para TVWS: onde é feita uma análise SWOT para o uso dos TVWS, dando direções/ações a serem seguidas para que o seu uso se torne efetivo; e um estudo tecno-econômico considerando como cenário uma típica cidade Europeia, onde se mostram as possíveis poupanças monetárias que os operadores conseguem obter ao optarem pelo uso dos TVWS num mercado flexível; (b) Comunicações D2D: onde uma técnica de descoberta de vizinhos para comunicações D2D é proposta, primeiro para uma única célula e mais tarde estendida para o cenário multi-celular; e um algoritmo de mitigação de interferência baseado na seleção inteligente da banda Ascendente (DL) ou Descendente (UL) a ser reusada pelas comunicações D2D que acontecem na rede celular. Um sumário das principais conclusões é o seguinte: (a) Os defensores dos TVWS devem-se focar na promoção do mercado secundário de espectro electromagnético, onde através da correta implementação de politicas de proteção contra a interferência no broker de espectro e na base de dados, os usuários primário são protegidos contra a interferência; (b) Um operador consegue recuperar o seu investimento aproximadamente um ano antes se ele optar pelo desenvolvimento da rede seguindo um mercado secundário de espectro com a banda adicional de TVWS, em vez do mercado tradicional; (c) Com a técnica proposta de descoberta de vizinhos, o tempo de descoberta por usuário é significativamente reduzido; e a potência consumida nesse processo é também ela reduzida porque o maior processamento é feito na Estação Rádio Base (BS), enquanto que o usuário só precisa de se certificar que a comunicação direta é possível; (d) A seleção de banda, embora seja um conceito simples, melhora os ganhos das comunicações celulares e limita os das comunicações D2D, providenciando um compromisso entre a performance do sistema e a mitigação de interferência.

From a fresh data science perspective, this thesis discusses the prediction of coronary artery disease based on Single-Nucleotide Polymorphisms (SNPs) from the Ontario Heart Genomics Study (OHGS). First, the thesis explains the k-Nearest Neighbour (k-NN) and Random Forest learning algorithms, and includes a complete proof that k-NN is universally consistent in finite dimensional normed vector spaces. Second, the thesis introduces two dimensionality reduction techniques: Random Projections and a new method termed Mass Transportation Distance (MTD) Feature Selection. Then, this thesis compares the performance of Random Projections with k-NN against MTD Feature Selection and Random Forest for predicting artery disease. Results demonstrate that MTD Feature Selection with Random Forest is superior to Random Projections and k-NN. Random Forest is able to obtain an accuracy of 0.6660 and an area under the ROC curve of 0.8562 on the OHGS dataset, when 3335 SNPs are selected by MTD Feature Selection for classification. This area is considerably better than the previous high score of 0.608 obtained by Davies et al. in 2010 on the same dataset.

Artificial intelligence has become a hot topic in the past couple of years because of its potential of solving problems. The most used subset of artificial intelligence today is machine learning, which is essentially the way a machine can learn to do tasks without getting any explicit instructions. A problem that has historically been solved by common knowledge and experience is the planning of bus transportation, which has been prone to mistakes. This thesis investigates how to extract the key features of a raw dataset and if a couple of machine learning algorithms can be applied to predict and plan the public bus transportation, while also considering the weather conditions. By using a pre-processing method to extract the features before creating and evaluating an k-nearest neighbors model as well as an artificial neural network model, predicting the passenger count on a given route could help planning of the bus transportation. The outcome of the thesis was that the feature extraction was successful, and both models could successfully predict the passenger count based on normal conditions. However, in extreme conditions such as the pandemic during 2020, the models could not be proven to successfully predict the passenger count nor being used to plan the bus transportation.

Questo studio tratta l'implementazione di due algoritmi di Machine Learning per il riconoscimento di tre attività motorie: camminata, corsa e andare in bicicletta. Mediante l'applicazione per smartphone Matlab Mobile sono state acquisii profili di accelerazione da tre utenti, due usati in fase di Test ed uno usato in fase di Training degli algoritmi. Si sono classificate tali attività estraendo delle opportune feature di interesse alla classificazione, in particolare la deviazione standard del''accelerazione che si è rivelata una buona scelta per discriminare le attività motorie. Gli algoritmi di classificazione testati sono K-Means e Nearest Neighbour e hanno dimostrato, tramite la matrice di confusione, di saper riconoscere se un utente sta svolgendo le attività citate rispettivamente nel 95,6% e 99.6% dei casi.

This work deals with the diagnosis of Parkinson's disease by analyzing the speech signal. At the beginning of this work there is described speech signal production. The following is a description of the speech signal analysis, its preparation and subsequent feature extraction. Next there is described Parkinson's disease and change of the speech signal by this disability. The following describes the symptoms, which are used for the diagnosis of Parkinson's disease (FCR, VSA, VOT, etc.). Another part of the work deals with the selection and reduction symptoms using the learning algorithms (SVM, ANN, k-NN) and their subsequent evaluation. In the last part of the thesis is described a program to count symptoms. Further is described selection and the end evaluated all the result.

Information on the location of mobile nodes in Mobile Ad-hoc Networks (MANETs) has the potential to significantly improve network performance. This thesis uses node location information to develop new techniques for route discovery in on-demand routing protocols such as the Ad-hoc On-Demand Distance Vector (AODV), thus making an important contribution to enhancing the experience of using mobile networks. A Candidate Neighbours to Rebroadcast the Route Request (CNRR) approach has been proposed to reduce the deleterious impact, known as the broadcast storm, of RREQ packets flooding in traditional on-demand routing protocols. The main concept behind CNRR is specifying a set of neighbours which will rebroadcast the received RREQ. This is a departure from the traditional approach of all receiving nodes rebroadcasting RREQs and has the effect of reducing the problem of redundancy from which mobile networks suffer. The proposed protocol has been developed in two phases: Closest-CNRR and Furthest-CNRR. The simulation results show that the proposed algorithms have a significant effect as they reduce the routing overhead of the AODV protocol by up to 28% compared to the C-CNRR, and by up to 17.5% compared to the F-CNRR. Notably, the proposed algorithms simultaneously achieve better throughput and less data dropping. The Link Stability and Energy Aware protocol (LSEA) has been developed to reduce the overhead while increasing network lifetimes. The LSEA helps to control the global dissemination of RREQs in the network by eliminating those nodes that have a residual energy level below a specific threshold value from participation in end-to-end routes. The proposed LSEA protocol significantly increases network lifetimes by up to 19% compared with other on-demand routing protocols while still managing to obtain the same packet delivery ratio and network throughput levels. Furthermore, merging the LSEA and CNRR concepts has the great advantage of reducing the dissemination of RREQs in the network without loss of reachability among the nodes. This increases network lifetimes, reduces the overhead and increases the amount of data sent and received. Accordingly, a Position-based Selective Neighbour (PSN) approach has been proposed which combines the advantages of zoning and link stability. The results show that the proposed technique has notable advantages over both the AODV and MAAODV as it improves delivery ratios by 24.6% and 18.8%, respectively.

This thesis deals with a comparison of classification methods. At first, these classification methods based on machine learning are described, then a classifier comparison system is designed and implemented. This thesis also describes some classification tasks and datasets on which the designed system will be tested. The evaluation of classification tasks is done according to standard metrics. In this thesis is presented design and implementation of a classifier that is based on the principle of evolutionary algorithms.

Combining the predictions of a set of classifiers has been shown to be an effective way to create composite classifiers that are more accurate than any of the component classifiers. Increased accuracy has been shown in a variety of real-world applications, ranging from protein sequence identification to determining the fat content of ground meat. Despite such individual successes, the answers are not known to fundamental questions about classifier combination, such as "Can classifiers from any given model class be combined to create a composite classifier with higher accuracy?" or "Is it possible to increase the accuracy of a given classifier by combining its predictions with those of only a small number of other classifiers?". The goal of this dissertation is to provide answers to these and closely related questions with respect to a particular model class, the class of nearest neighbor classifiers. We undertake the first study that investigates in depth the combination of nearest neighbor classifiers. Although previous research has questioned the utility of combining nearest neighbor classifiers, we introduce algorithms that combine a small number of component nearest neighbor classifiers, where each of the components stores a small number of prototypical instances. In a variety of domains, we show that these algorithms yield composite classifiers that are more accurate than a nearest neighbor classifier that stores all training instances as prototypes. The research presented in this dissertation also extends previous work on prototype selection for an independent nearest neighbor classifier. We show that in many domains, storing a very small number of prototypes can provide classification accuracy greater than or equal to that of a nearest neighbor classifier that stores all training instances. We extend previous work by demonstrating that algorithms that rely primarily on random sampling can effectively choose a small number of prototypes.

碩士
國立臺灣科技大學
資訊工程系
99
Data reduction is to extract a subset from a dataset. The advantage of data reduction is decreasing the requirement of storage. Using the subset as training data is possible to maintain classification accuracy; sometimes, it can be further improved because of eliminating noises. The key is how to choose representative samples while ignoring noises at the same time. Many instance selection algorithms are based on Nearest Neighbor decision rule (NN). Some of these algorithms select samples based on two strategies, incremental and decremental. The first type of algorithms selects some instances as samples and iteratively adds instances which do not have the same class label with their nearest sample to the sample set. The second type of algorithms gradually removes instances based on its own strategies. However, we propose an algorithm based on Reverse Nearest Neighbor (RNN), called Reverse Nearest Neighbor Reduction (RNNR). RNNR selects samples which can represent other instances in the same class. In addition, RNNR does not need to iteratively scan a dataset which takes much processing time. Experimental results show that RNNR generally achieves higher accuracy, selects fewer samples and takes less processing time than comparators.

Probabilistic Roadmap Methods (PRMs) are one of the most used classes of motion planning methods. These sampling-based methods generate robot configurations (nodes) and then connect them to form a graph (roadmap) containing representative feasible pathways. A key step in PRM roadmap construction involves identifying a set of candidate neighbors for each node. Traditionally, these candidates are chosen to be the k-closest nodes based on a given distance metric. This work proposes a new neighbor selection policy called LocalRand(k, k'), that first computes the k' closest nodes to a specified node and then selects k of those nodes at random. Intuitively, LocalRand attempts to benefit from random sampling while maintaining the higher levels of local planner success inherent to selecting more local neighbors. A methodology for selecting the parameters k and k' is provided, and an experimental comparison for both rigid and articulated robots show that LocalRand results in roadmaps that are better connected than the traditional k-closest or a purely random neighbor selection policy. The cost required to achieve these results is shown to be comparable to the cost of k-closest.

碩士
元智大學
工業工程與管理學系
107
In this study, we proposed Variable Neighborhood Search(VNS) algorithm to solve the problem of feature selection.By searching different solutions, we uses the k-Nearest Neighbors algorithm(kNN) to evaluate the classification performance of the solution.It is expected to pick out a subset of feature that can effectively classify the data with a small number of features. This study first conducts a series of parametric experiments on the nearest neighbor method (KNN) to find out the parameter set that can have good classification performance in most data sets. Then we tested the algorithm in data sets with different sizes. The experimental results compared with previous studies shows that the proposed method can achieve similar or even better classification performance in most data sets than in previous studies.

碩士
國立中興大學
農藝學系所
94
Microarray technology has become a valuable tool for studying gene expression in recent years. The main difference between microarray and traditional methods is that microarray can measure thousands of genes at the same time. In the past, researchers always used parametric statistical methods to find the significant genes. However, microarray data often cannot obey some assumptions of parametric statistical methods, and type I error would be over expanded while each gene was tested for significance. Therefore, this research was expected to find a variable selection method without assumptions restriction to reduce the dimension of the data set. After using the proposed method, biologists can select the relevant genes according to the sub-gene set. In this study, adaptive genetic algorithms / k-nearest neighbors (AGA / KNN) was used to reduce the dimension of the data set, and it was based on genetic algorithms / k-nearest neighbors (GA / KNN) which was first described by Li et al.(2001a). Although AGA and KNN were well-developed, AGA / KNN was first used to analyze the microarray data. Since AGA was a machine learning tool and KNN was a nonparametric discrimination analysis, both of them could be used without assumptions restriction. There are three main differences between AGA/KNN and GA / KNN. Firstly, the encoding has become binary code, and each string included all genes. Secondly, the adaptive probabilities of crossover and mutation were added. Finally, the extinction and immigration strategy was added. Since GA can just find the near optimal solution, the best string of each run is often not the same. Here, AGA / KNN was repeated by many runs to solve that problem. Thus, lots of the best strings were saved. The frequency of gene was computed by those strings to reduce the dimension of the data set. In this study, an original colon data which is a high-density oligonucleotide chip (Alon et al., 1999) was analyzed. In addition, mice apo AI data which is a cDNA chip (Callow et al., 2000) was also used to compare the ability of gene selection of AGA / KNN and GA / KNN. Based on the results, it was found that AGA / KNN and GA / KNN could reduce the dimension of the data set and all samples could be classified correctly. But the accuracy of AGA / KNN was higher than that of GA / KNN, and it only took half CPU time of GA / KNN. Therefore, it was claimed that the performance of AGA / KNN should not be worse than that of GA / KNN. Finally, we suggested that when AGA / KNN was employed to analyze the microarray data, the top 50 and up to 100 most frequent genes were selected after AGA / KNN were repeated about 100 runs. Those selected genes should include relevant genes, and those selected genes could classify sample correctly.

Bike-sharing systems (BSS) are becoming very popular and, consequently, their management is becoming more complex. The main objective of this dissertation is the development of a decision support service for Bewegen bike-sharing systems applying machine learning (ML) methods. An additional objective is the development of an appropriate mechanism for systematic data collection, required in the development and test of the ML methods. The decision support service has two goals. The first goal is the prediction of the number of bikes in each station 30 minutes ahead of time, to be provided to the bike-sharing system clients. The second goal is the prediction of the number of bikes in each station 24 hours ahead of time, to be provided to the bike-sharing operators when deciding how to redistribute bikes among the different stations. In order to reach these two goals, two ML approaches were implemented: a neural network (NN) model and a k-nearest neighbour (k-NN) algorithm. The tests have shown that the NN algorithms provide better prediction results on both goals. The prediction algorithms were trained and tested with collected historical data from one of the Bewegen's BSS from 1 of January, 2019 until 30 of April, 2019.
Os sistemas de bike-sharing estão a tornar-se cada vez mais populares e a sua gestão mais complexa. O objetivo principal desta dissertação é o desenvolvimento de um serviço de suporte de decisão, baseado em métodos de aprendizagem automática, para os sistemas de bikesharing da empresa Bewegen. Um objetivo secundário é o desenvolvimento de um mecanismo de recolha sistemática de dados de utilização do sistema, necessários ao desenvolvimento e teste dos métodos de aprendizagem automática. O serviço de suporte de decisão tem dois objetivos. O primeiro objetivo é a previsão do número de bicicletas em cada estação com 30 minutos de antecedência, informação esta a disponilizar aos clientes do sistema de bike-sharing. O segundo objetivo é a previsão do número de bicicletas em cada estacão com 24 horas de antecedência, informação esta a disponilizar aos operadores do sistema no planeamento da distribuição das bicicletas pelas diferentes estacões. Para cumprir com estes objetivos foram implementados dois algoritmos de aprendizagem automática: uma rede neuronal e um algoritmo k-nearest neighbour. Os testes realizados mostram que os algoritmos baseados em redes neuronais obtém melhor desempenho nos dois objectivos. Os dados utilizados nos testes dos dois algoritmos são os dados históricos de um dos sistemas da Bewegen recolhidos desde 1 de janeiro de 2019 até 30 de abril de 2019.
Mestrado em Engenharia de Computadores e Telemática

Ο στόχος αυτής της εργασίας ήταν η αναγνώριση έξι βασικών κινήσεων του χεριού με χρήση δύο συστημάτων. Όντας θέμα διεπιστημονικού επιπέδου έγινε μελέτη της ανατομίας των μυών του πήχη, των βιοσημάτων, της μεθόδου της ηλεκτρομυογραφίας (ΗΜΓ) και μεθόδων αναγνώρισης προτύπων. Παράλληλα, το σήμα περιείχε αρκετό θόρυβο και έπρεπε να αναλυθεί, με χρήση του EMD, να εξαχθούν χαρακτηριστικά αλλά και να μειωθεί η διαστασιμότητά τους, με χρήση των RELIEF και PCA, για βελτίωση του ποσοστού επιτυχίας ταξινόμησης. Στο πρώτο μέρος γίνεται χρήση συστήματος ΗΜΓ της Delsys αρχικά σε ένα άτομο και στη συνέχεια σε έξι άτομα με το κατά μέσο όρο επιτυχημένης ταξινόμησης, για τις έξι αυτές κινήσεις, να αγγίζει ποσοστά άνω του 80%. Το δεύτερο μέρος περιλαμβάνει την κατασκευή αυτόνομου συστήματος ΗΜΓ με χρήση του Arduino μικροελεγκτή, αισθητήρων ΗΜΓ και ηλεκτροδίων, τα οποία είναι τοποθετημένα σε ένα ελαστικό γάντι. Τα αποτελέσματα ταξινόμησης σε αυτή την περίπτωση αγγίζουν το 75%.
The aim of this work was to identify six basic movements of the hand using two systems. Being an interdisciplinary topic, there has been conducted studying in the anatomy of forearm muscles, biosignals, the method of electromyography (EMG) and methods of pattern recognition. Moreover, the signal contained enough noise and had to be analyzed, using EMD, to extract features and to reduce its dimensionality, using RELIEF and PCA, to improve the success rate of classification. The first part uses an EMG system of Delsys initially for an individual and then for six people with the average successful classification, for these six movements at rates of over 80%. The second part involves the construction of an autonomous system EMG using an Arduino microcontroller, EMG sensors and electrodes, which are arranged in an elastic glove. Classification results in this case reached 75% of success.

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies
This work presents a Sentinel-2 based exploratory work ow for the estimation of Above Ground Biomass (AGB) and Carbon Sequestration (CS) in a subtropical forest. In the last decades, remote sensing-based studies on AGB have been widely investigated alongside with a variety of sensors, features and Machine Learning (ML) algorithms. Up-to-date and reliable mapping of such measures have been increasingly required by international commitments under the climate convention as well as by sustainable forest management practices. The proposed approach consists of 5 major steps: 1) generation of several Vegetation Indices (VI), biophysical parameters and texture measures; 2) feature selection with Mean Decrease in Impurity (MDI), Mean Decrease in Accuracy (MDA), L1 Regularization (LASSO), and Principal Component Analysis (PCA); 3) feature selection testing with k-Nearest Neighbour (kNN), Random Forest (RF), Extreme Gradient Boosting (XGB), and Arti cial Neural Network (ANN); 4) hyper-parameters ne-tuning with Grid Search, Random Search and Bayesian Optimization; and nally, 5) model explanation with the SHapley Additive exPlanations (SHAP) package, which to this day has not been investigated in the context of AGB mapping. The following results were obtained: 1) MDI was chosen as the best performing feature selection method by the XGB and the Deep Neural Network (DNN), MDA was chosen by the RF and the kNN, while LASSO was chosen by the Shallow Neural Network (SNN) and the Linear Neural Network (LNN); 2) before hyper-parameters optimization, the Deep Neural Network (DNN) yielded the best performance with a Root Mean Squared Error (RMSE) of 42.30 t=ha; 3) after hyper-parameters ne-tuning with Bayesian Optimization, the XGB model yielded the best performance with a RMSE of 37.79 t=ha; 4) model explanation with SHAP allowed for a deeper understanding of the features impact on the model predictions. Finally, the predicted AGB throughout the study area showed an average value of 83 t=ha, ranging from 0 t=ha to 346.56 t=ha. The related CS was estimated by using a conversion factor of 0.47.