Made available in DSpace on 2017-07-21T14:19:40Z (GMT). No. of bitstreams: 1 Marinaldo Gaspareto.pdf: 1456191 bytes, checksum: ffaf0b449c6b9d107bdf1946a4619315 (MD5) Previous issue date: 2013-09-10 / A problem regarding the implementation of navigation systems for autonomous moving robots is to detect the objects of interest and obstacles which are in the environment. This study considers the detection of walls / low walls of agricultural greenhouses in digital images obtained without illumination control. The proposed approach employs techniques of digital image processing and digital classification to detect the object of interest. The classifier has been developed digital type Naive Bayes. Two important issues when employing classification methods in computer vision is the accuracy of the classifier and the complexity of computing time. The selection of attributes descriptors that comprise a classifier has great impact on these two factors, generally the fewer attributes are required, the lower the computational cost. Regarding it, this study compared the performance of two methods of feature selection based on principal component analysis, named B2 and B4 in two cases. In the first scenario the feature selection was conducted on all the data extracted from all images. The second selection was performed for images grouped by similarity. After selection, the selected attributes for each approach was used to construct the type Naive Bayes classifier with 12, 17, 22 and 27 input variables. The results indicate that the grouping of images is useful when: (a) the distance from the center of the group to the center of the original database exceeds a threshold and (b) a correlation among the descriptors variables and the target variable is greater than in the group as a whole complete data. Keywords: Greenhouses, Autonomous navigation, Selection attributes, Naive Bayes classifiers. / Um problema relativo à implementação de sistemas de navegação para robôs autônomos móveis é a detecção dos objetos de interesse e dos obstáculos que estão no ambiente. Este trabalho considera a detecção das paredes/muretas de estufas agrícolas em imagens digitais adquiridas sem controle de iluminação. A abordagem proposta emprega técnicas de processamento digital de imagens e classificação digital para detectar o objeto de interesse. O classificador digital desenvolvido foi do tipo Naive Bayes. Duas questões importantes quando do emprego de métodos de classificação em visão computacional são a acurácia do classificador e a complexidade de tempo de computação. A seleção dos atributos descritores que compõem um classificador tem grande impacto sobre estes dois fatores, de um modo geral, quanto menos atributos forem necessários, menor o custo computacional. Considerando isso, este trabalho comparou o desempenho de dois métodos de seleção de atributos baseados na análise de componentes principais, chamados B2 e B4 em duas situações. Na primeira situação, a seleção de atributos foi realizada sobre o conjunto dos dados extraídos de todas as imagens. Na segunda, a seleção foi realizada para imagens agrupadas por similaridade. Após a seleção, os atributos selecionados em cada uma das abordagens foram usados para construir classificadores do tipo Naive Bayes com 12, 17, 22 e 27 variáveis de entrada. Os resultados indicam que o agrupamento de imagens é útil quando: (a) a distância do centro do grupo ao centro da base original ultrapassa um limiar e (b) a correlação entre as variáveis descritoras e a variável meta é maior no grupo do que no conjunto completo de dados.

The aim of this study is to evaluate feasibility of using Bayes classifiers for predicting corporate bankruptcies. The results obtain show that Bayes classifiers do reach comparable results to then more commonly used methods such the logistic regression and the decision trees. The comparison has been carried out based on Czech and Polish data sets. The overall accuracy rate of these so called naive Bayes classifiers, using entropic discretization along with the hybrid pre-selection of the explanatory attributes, reaches 77.19 % for the Czech dataset and 79.76 % for the Polish set respectively. The AUC values for these data sets are 0.81 and 0.87. The results obtained for the Polish data set have been compared to the already published articles by Tsai (2009) and Wang et al. (2014) who applied different classification algorithms. The method proposed in my study, when compared to the above earlier works, comes out as quite successful. The thesis also includes comparing various approaches as regards the discretisation of numerical attributes and selecting the relevant explanatory attributes. These are the key issues for increasing performance of the naive Bayes classifiers

This study examines different methods of language identification for the languages Meänkieli, Kven, and Finnish. The methods explored are two n-gram-based classifiers; Naive Bayes and TextCat and one word embedding-based classifier; fastText. These models were trained on approximately 100.000 sentences taken from the three languages and further divided into four separate datasets to examine how data availability impacts the final performance of the trained models. The study found that the best model for the examined dataset was the fastText classifier, but for languages with less available material a naive Bayes classifier might be more appropriate. / Denna studie utforskar olika metoder av språkidentifering för språken meänkieli, kvänska och finska. Två metoder baserade på n-gram undersöks; naive Bayes och TextCat samt en metod med ordinbäddningar; fastText. Dessa modeller tränades på sammanlagt 100 000 meningar taget från dessa tre språk och delades vidare in i fyra delmängder för att utvärdera hur stor inverkan storleken av träningsdata har på de tränade modellerna. Studien fann att den bästa implementationen utifrån den undersökta datamängden var fastText, medans språk med färre resurser skulle förmodligen gynnas bättre av en språkidentifering byggd med en naive Bayes klassifierare.