Machine learning (ML) techniques

Top PDF Machine learning (ML) techniques:

ISOLATED SPEECH RECOGNITION SYSTEM FOR TAMIL LANGUAGE USING STATISTICAL PATTERN MATCHING AND MACHINE LEARNING TECHNIQUES

ISOLATED SPEECH RECOGNITION SYSTEM FOR TAMIL LANGUAGE USING STATISTICAL PATTERN MATCHING AND MACHINE LEARNING TECHNIQUES

Based on the experiments, it was found that the HMM and DTW has provided the better results followed by MLP and SVM when compared to the other methods. High recognition rate was achieved for word/utterance level performance using HMM. By using HMM, 100% accuracy was obtained for all the words during training process. For test data also, HMM provided 100% accuracy for 11 words out of fifteen words for all the speakers involved. Next to HMM, DTW was also obtained 100% accuracy for 10 words out of fifteen words during testing process. Next to these two methods, MLP and SVM have offered considerable results for both training and test data. Like HMM, MLP has also reached a great improvement by offering 100% results for training data. Based on the time factor, it was observed that the statistical approaches are time consuming when compared to machine learning techniques. Processing time taken for DTW also found to be high when compared to MLP, SVM and DT. The decision tree algorithm has taken very less processing time, but produced very less accuracy for the above developed system. Based on the utterance level performance, the digit seven (D7) has achieved 100% accuracy for all the techniques and speakers involved in the experiments. Next to D7, the D6, D8, W4 and W5 gave high accuracy for all the speakers enrolled in the study. It was observed from the experimental results that, the performance of the above system was found to be good when compared with the results achieved for the existing system discussed here. The average word recognition accuracy achieved for test data is 95.83%, 97.92%, 86.67%, 92.50%, 90.83%, 63.33% for DTW, HMM, GMM, MLP, SVM and Decision tree algorithms respectively. Particularly, 97.92% is achieved by HMM which is higher than the existing results. By considering the above experiments and analysis factor, HMM and DTW followed by MLP and SVM gives better recognition rate for the above developed system. Specifically, statistical approach improves word level performance and machine learning approaches works better for speaker independent applications.
Mostrar mais

16 Ler mais

Twitter Sentiment Analysis of Movie Reviews using Machine Learning Techniques.

Twitter Sentiment Analysis of Movie Reviews using Machine Learning Techniques.

In general, various symbolic techniques and machine learning techniques are used to analyze the sentiment from the twitter data. So in another way we can say that a sentiment analysis is a system or model that takes the documents that analyzed the input, and generates a detailed document summarizing the opinions of the given input document. In the first step pre-processing is done. In the pre-processing we are removing the stop words, white spaces, repeating words, emoticons and #hash tags.

7 Ler mais

Comparative Performance Analysis of Machine Learning Techniques for Software Bug Detection

Comparative Performance Analysis of Machine Learning Techniques for Software Bug Detection

Lessmann et al. [12] proposed a novel framework for software defect prediction by benchmarking classification algorithms on different datasets and observed that their selected classification methods provide good prediction accuracy and supports the metrics based classification. The results of the experiments showed that there is no significant difference in the performance of different classification algorithms. The study did not cover all machine learning techniques for software bug prediction. Sharma and Jain [13] explored the WEKA approach for different classification algorithms but they did not explore them for software bug prediction. Kaur and Pallavi [14] explored the different data mining techniques for software bug prediction but did not provide the comparative performance analysis of techniques. Wang et al. [15] provided a comparative study of only ensemble classifiers for software bug prediction. Most of the existed studies on software defect prediction are limited in performing comparative analysis of all the methods of machine learning. Some of them used few methods and provides the comparison between them and others just discussed or proposed a method based on existing machine learning techniques by extending them [16, 17, 18].
Mostrar mais

9 Ler mais

MACHINE LEARNING TECHNIQUES USED IN BIG DATA

MACHINE LEARNING TECHNIQUES USED IN BIG DATA

Cluster analysis. Cluster analysis splits the data in different groups, called clusters, which are relevant, helpful, or both. If the relevant groups are the purpose, then the clusters should catch the natural structure of the information. Sometimes, cluster analysis represents a beginning point for another scopes, like summarization. Cluster analysis was one of the most important techniques, applied in different areas of study, like social sciences, natural sciences, medicine or businesses [17]. There are two main types of clustering: partitional clustering (Figure 3) and hierarchical clustering. The partitional clustering just divides the dataset objects into distinct subsets (clusters), as every object stands in only one subset. Thus, every collecting of clusters represents a partitional clustering.If it is allowed the clusters to have sub- clusters, then we are talking about hierarchical clustering that represent a group of nested clusters, represented as a tree. Every node, excepting the leaves, is the unification of its children nodes, so the root represents the initial set of objects[17]. Sometimes, the leaves represent sets of a single object. Cluster analysis does not represent itself an algorithm, but represent the general task, which should be accomplished. Examples of clustering algorithms are: k-means - for centroid models, density-based spatial clustering of applications with noise (DBSCAN) orordering points to identify the clustering structure (OPTICS) – for density models, biclustering or co-clustering – for subspace models, etc. [18], [19].
Mostrar mais

6 Ler mais

Analysis of Temporal Variations in Dermoscopy Images of Pigmented Skin Lesions by Machine Learning Techniques

Analysis of Temporal Variations in Dermoscopy Images of Pigmented Skin Lesions by Machine Learning Techniques

In ANNs the connections are established between layers of input and output elements with different associated weights that are adjusted during the learning phase in order to predict the correct class output of the input samples. A widely used ANN architecture is the multilayer perceptron (MLP) that allows more than one layer of processing (hidden layer) and the back- propagation learning algorithm that uses backward and forward processes to adjust the weight values of the connections. These classifiers have some disadvantages related to training time that may be very long depending on training set size.
Mostrar mais

64 Ler mais

Machine-learning techniques and short-term combination forecasting of industrial production

Machine-learning techniques and short-term combination forecasting of industrial production

The industrial production index is an important measure when assessing the eco- nomic scenario and is used by both state and private institutions for decision-making. In Brazil, it’s published every month by the Brazilian Institute of Geography and Statistics (IBGE, 2018), but not until approximately 35 days after the reference period. Because of the importance of this variable in any analysis of the economic scenario, a vari- ety of techniques are used in the process of forecasting it. Examples include dynamic regressions (Madsen, 1993); structural models (Thury & Witt, 1998); ARIMA, VAR, structural VAR, VECM and dynamic factor models (Bodo et al., 2000; Hollauer et al., 2008; Bulligan et al., 2010; Costantini, 2013); univariate and multivariate singular spec- trum analysis (Hassani et al., 2013); mixed-frequency models (Seong et al., 2013); and Markov regime models (Bazzi et al., 2017). Techniques that combine forecasts made using different methods into a single more accurate forecast have also been used (Elliott et al., 2006; Hollauer et al., 2008; Bulligan et al., 2010; Vasnev et al., 2013; Berge, 2015) and have proven effective in reducing forecast error.
Mostrar mais

10 Ler mais

Customer Clustering in the Insurance Sector by Means of Unsupervised Machine Learning Techniques

Customer Clustering in the Insurance Sector by Means of Unsupervised Machine Learning Techniques

10 The second divisive algorithm that should be presented is bisecting k-means. The bisecting k-means clustering algorithm tries to combine the time efficiency of the k-means algorithm with the high quality results from hierarchical clustering (Steinbach, Karypis, & Kumar, 2000). For now it is enough to know that the k-means algorithm divides a chosen cluster into two sub clusters. A detailed discussion of the k-means algorithm is given in K-Means Algorithm. The algorithm starts with all observations in a single cluster. This cluster is split into two sub clusters by using the standard k- means algorithm. In order to obtain the best split of a chosen cluster, the k-means algorithm is initialized a number of times. The iteration that results with the lowest average pairwise distance of within the sub clusters is chosen to perform the split. This procedure is repeated until the predetermined number of clusters is reached. Steinbach et al. (2000) note, that there is only little difference between the possible methods of selecting a cluster to split in each iteration. These methods can be based on the techniques that are introduced in Internal Validation Techniques.
Mostrar mais

122 Ler mais

Exploration of machine learning techniques for automatic optical inspection

Exploration of machine learning techniques for automatic optical inspection

The obtained data from each sensor is used to calculate a Feature Vector [27]. A Feature Vector consists of an n-dimensional vector of numerical features that represent some object, so in the case of images, these vectors represent their pixels. The score obtained results in the calculation between the calculated vector’s proximity at the time the supposed user is in contact with the sensor and the vector that was generated when the model was created with the user’s identity [27]. Techniques, such as logistic regression are used to combine the results, if they come from different sensors, since several biometric indicators are being evaluated at the same time and, as such, this combination of results is made so that the truth can be affirmed of the identity that is being presented.
Mostrar mais

93 Ler mais

Project Work presented as partial requirement for obtaining the Master’s degree in Information Management

Project Work presented as partial requirement for obtaining the Master’s degree in Information Management

Machine learning techniques are used in variety of industries, education, medicine, chemistry, and a lot of other science fields. Machine learning approaches are applied to solve real-world business problems, like found in financial services, transportation, or in marketing and sales perspectives. For the banks and other financial businesses, machine learning is used to identify investment opportunities, clients with high-risk profiles, when to trade or prevent fraud. In transportation industry, like in delivery companies or public transportation, machine learning approaches are applied to find inherent patterns and trends for making routes more efficient and for predicting potential problems on the road. In the marketing and sales perspectives, machine learning methods are used for customer churn prediction, target marketing or it is also used to analyse buying behaviour of customers and make promotions based on analysis. (Linoff & Berry, 2011; "Machine Learning: What it is and why it matters", 2017)
Mostrar mais

129 Ler mais

Active learning to understand infectious disease models and improve policy making.

Active learning to understand infectious disease models and improve policy making.

Modeling plays a major role in policy making, especially for infectious disease interventions but such models can be complex and computationally intensive. A more systematic exploration is needed to gain a thorough systems understanding. We present an active learning approach based on machine learning techniques as iterative surrogate modeling and model-guided experimentation to systematically analyze both common and edge manifestations of complex model runs. Symbolic regression is used for nonlinear response surface modeling with automatic feature selection. First, we illustrate our approach using an individual-based model for influenza vaccination. After optimizing the parameter space, we observe an inverse relationship between vaccination coverage and cumulative attack rate reinforced by herd immunity. Second, we demonstrate the use of surrogate modeling techniques on input-response data from a deterministic dynamic model, which was designed to explore the cost-effectiveness of varicella-zoster virus vaccination. We use symbolic regression to handle high dimensionality and correlated inputs and to identify the most influential variables. Provided insight is used to focus research, reduce dimensionality and decrease decision uncertainty. We conclude that active learning is needed to fully understand complex systems behavior. Surrogate models can be readily explored at no computational expense, and can also be used as emulator to improve rapid policy making in various settings.
Mostrar mais

10 Ler mais

Developing a Neural Network based Index for Sentiment Classification

Developing a Neural Network based Index for Sentiment Classification

Abstract—Recognizing emotion is extremely important for some text-based communication tools such as blogs. On commercial blogs, bloggers’ negative comments or evaluations of products spread quickly in the cyber space. These negative comments are often harmful to enterprises and might result in great damage. Recently, researchers have paid much attention on sentiment classification to efficiently identify customers’ negative emotions for helping companies to carefully response customers’ comments. Following this trend, this study proposed a Neural Network (NN) based index which combines the advantages of machine learning techniques and information retrieval (semantic orientation indexes) to help companies detecting harmfully negative bloggers’ comments quickly and effectively. Experimental results indicated that our proposed NN based index outperforms traditional approaches, including Back-Propagation neural network (BPN) and several semantic orientation indexes.
Mostrar mais

6 Ler mais

Machine learning and its applications to biology.

Machine learning and its applications to biology.

The history of relations between biology and the field of machine learning is long and complex. An early technique [1] for machine learning called the perceptron constituted an attempt to model actual neuronal behavior, and the field of artificial neural network (ANN) design emerged from this attempt. Early work on the analysis of translation initiation sequences [2] employed the perceptron to define criteria for start sites in Escherichia coli. Further artificial neural network architectures such as the adaptive resonance theory (ART) [3] and neocognitron [4] were inspired from the organization of the visual nervous system. In the intervening years, the flexibility of machine learning techniques has grown along with mathematical frameworks for measuring their reliability, and it is natural to hope that machine learning methods will improve the efficiency of discovery and understanding in the mounting volume and complexity of biological data.
Mostrar mais

11 Ler mais

Modelos complexos de predição aplicados na educação

Modelos complexos de predição aplicados na educação

The classification trees have two main basic tuning parameters (for more fine grained tuning parameters see Breiman, Friedman, Olshen & Stone, 1984): 1) the number of features used in the prediction , and 2) the complexity of the tree, which is the number of possible terminal nodes 𝛼| |. Geurts, Irrthum and Wehenkel (2009) argue that classification trees are among the most popular algorithms of Machine Learning due to three main characteristics: interpretability, flexibility and ease of use. Interpretability means that the model constructed to map the feature space into the output space is easy to understand, since it is a roadmap of if-then rules. James, Witten, Hastie and Tibshirani (2013) points that the tree models are easier to explain to people than linear regression, since it mirrors more the human decision- making then other predictive models. Flexibility means that the tree techniques are applicable to a wide range of problems, handles different kind of variables (including nominal, ordinal, interval and ratio scales), are non-parametric techniques, does not make any assumption regarding normality, linearity or independency and can be applied in datasets with a large p low n characteristic (Geurts, et al., 2009). Furthermore, it is sensible to the impact of additional variables to the model, being especially relevant to the study of incremental validity. Finally, the ease of use means that the tree based techniques are computationally simple, yet powerful.
Mostrar mais

139 Ler mais

Deteção de fraude em telecomunicações através de machine learning

Deteção de fraude em telecomunicações através de machine learning

Em concreto, esta dissertação debruçou-se sobre uma base de dados de uma opera- dora, com informação sobre chamadas recebidas numa gateway, tendo por objetivo a identificação de fraudes do tipo bypass e wangiri. Em primeiro lugar, foi desen- volvida uma análise exploratória com base em análises estatísticas, para melhor conhecimento dos dados, tendo sido criados novos atributos para ajudarem os mo- delos. Um atributo que teve um papel fundamental nesta dissertação foi a Range, que se baseia no agrupamento de números telefónicos, tendo em conta a variação dos últimos dígitos dos números. Posteriormente, foram desenvolvidos modelos de machine learning sem supervisão: PCA, autoencoder e LSTM autoencoder. Uma das conclusões deste trabalho é a de que os bons resultados produzidos pelo modelo PCA, sugerem que a não fraude possa ser um problema linear, apesar de produzir uma percentagem elevada de outliers. Os modelos de autoencoder por si só não produziram tão bons resultados, mas após aplicação de filtros baseados em scores (de forma a tentar quantificar a não linearidade dos dados), observou-se uma acentuada melhoria nos resultados. Os resultados preliminares obtidos com os modelos LSTM autoencoders sugerem que a sua capacidade de guardar dados em memória pode vir a produzir muito bons resultados.
Mostrar mais

110 Ler mais

Parsimonious Wavelet Kernel Extreme Learning Machine

Parsimonious Wavelet Kernel Extreme Learning Machine

In the current study, a parsimonious algorithm and wavelet technique are developed for KELM on the basis of the aforementioned analysis, the resultant model is referred to as parsimonious wavelet kernel extreme learning machine (PWKELM). Wavelet function benefits from multiscale interpolation and is also suitable for the local analysis and detection of transient signals. As a result, a wavelet expansion representation is compact and is easy to implement. Householder matrix is also used to orthogonalize the linear equation set in WKELM, and significant wavelet kernel functions are recruited iteratively. Thus, a sparse solution is established. Synthetic and real-world data sets are utilized in conducting experiments whose results confirm the effectiveness and feasibility of the proposed PWKELM.
Mostrar mais

8 Ler mais

A Heuristic Approach to the Disease Diagnose System Using Machine Learning Algorithms (An Expert Advisory System for EMU Bird’s Diseases)

A Heuristic Approach to the Disease Diagnose System Using Machine Learning Algorithms (An Expert Advisory System for EMU Bird’s Diseases)

Abstract--The paper deals with the concepts of expert system and data mining belongs to the Artificial Intelligence fields. The main task of expert system is to ratiocination, while the machine learning algorithm is to find the better optimal solution. This paper mainly focuses on diagnoses of the disease which is effected to the Emu bird by mechanism of Particle Swarm Optimization (PSO) algorithm and Artificial Bee Colony(ABC) algorithm. The decisive rules of database is mined and that could be applied in expert system. Thus, by applying optimization techniques resulting to best global optimized solution.
Mostrar mais

3 Ler mais

An Optimized Extreme Learning Machine for Epileptic Seizure Detection

An Optimized Extreme Learning Machine for Epileptic Seizure Detection

Abstract—In this paper, a novel learning framework for Single hidden Layer Feed forward Neural network (SLFN) called Optimized Extreme Learning Machine (OELM) is proposed for the classification of EEG signals with the emphasis on epileptic seizure detection. OELM is an effective learning algorithm of single-hidden layer feed-forward neural networks. It requires setting the number of hidden neurons and the activation function. Adjustment in the input weights and hidden layer’s biases are not needed during the implementation of the algorithm, and only one optimal solution is produced. This makes the OELM a valuable tool for the applications that need small response time and provide a good accuracy. The features such as energy, entropy, maximum value, minimum value, mean value and standard deviation of wavelet coefficients are used to represent the time frequency distribution of the EEG signals in each sub-band of the Wavelet Transformation. We have compared the proposed classifier with other traditional classifiers by evaluating it with the benchmark EEG dataset. It is found that the performance of the proposed OELM with Wavelet based statistical features is better in terms of training time and classification accuracy. An accuracy of 94% for classifying the epileptic EEG signals is achieved and needs less training time compared with SVM.
Mostrar mais

10 Ler mais

Adversarial machine learning: denial of services recognition

Adversarial machine learning: denial of services recognition

Zilong Lin et al. proposed IDSGAN [18], a Generative Adversarial Network (GAN) applica- tion to generate indiscriminate adversaries on the NSL-KDD dataset [51]. By training a Generator to produce adversarial malicious traffic from real malicious traffic, and a discriminator that dis- tinguished normal traffic from adversarial, following the Wesserstein GAN arquitecture [28], the authors were able to generate adversarial network traffic. The effectiveness of the attacks was tested against a multitude of classifiers, namely Support Vector Machine, Naive Bayes, Multi- layer Perceptron, Logistic Regression, Decision Tree, Random Forest and K-nearest neighbors. By applying this concept to the NSL-KDD dataset, the results showed that the detection rates of adversarials on all classifier dropped from over 70% on the original data to less than 1% for all classifiers and types of attacks. To make the attacks more realistic, the authors prevented the IDSGAN from modifying functional features of the attacks, which are features that would affect the performance of the intrusions, such as time-based features for Denial of Service attacks, or content based features for User-to-Root attacks. Howerever, they observe no notorious difference from the performance with the inclusion of functional features, proving thus the effectiveness of IDSGAN at creating adversarial examples, although there were no limitations to the amount of noise introduced by the technique.
Mostrar mais

89 Ler mais

Rev. bras. linguist. apl.  vol.11 número2

Rev. bras. linguist. apl. vol.11 número2

A second question concerns how frequency of occurrence and co- occurrence frequencies come into play in human classification behavior as compared to machine classification. For machine classification, we can easily count how often a linguistic element occurs, and how often it co-occurs with other elements. The success of machine classification in reproducing linguistic choice behavior suggests that probabilities of occurrence are somehow available to the human classifier. But is frequency of (co-)occurrence available to the human classifier in the same way as to the machine classifier? Simple frequency of occurrence information is often modeled by means of some ‘counter in the head’, implemented in cognitive models in the form of ‘resting activation levels’, as in the interactive activation models of McClelland and Rumelhart (1981); Coltheart, Rastle, Perry, Langdon, and Ziegler (2001); Van Heuven, Dijsktra, and Grainger (1998), in the form of frequency based rankings (MURRAY; FORSTER, 2004), as a unit’s verification time (LEVELT, ROELOFS; MEYER, 1999), or in the Bayesian approach of Norris, straightforwardly as a unit’s long-term a-priori probability (NORRIS, 2006; NORRIS; McQUEEN, 2008). A potential problem that arises in this context is that large numbers of such ‘counters in the head’ are required, not only for simple or complex words, but also for hundreds of millions of word n-grams, given recent experimental results indicating human sensitivity to n-gram frequency (ARNON; SNIDER, 2010; TREMBLAY; BAAYEN, 2010). Moreover, given the tendency of human memory to merge, or blend, previous experiences, it is rather unlikely that the human classifier has at its disposal exactly the same frequency information that we make available to our machine classifiers.
Mostrar mais

34 Ler mais

Aplicação de Algoritmos de Aprendizado de Máquina no Apoio a Elaboração de Planos Nutricionais / Application of Machine Learning Algorithms to Support the Development of Nutritional Plans

Aplicação de Algoritmos de Aprendizado de Máquina no Apoio a Elaboração de Planos Nutricionais / Application of Machine Learning Algorithms to Support the Development of Nutritional Plans

Machiori, Stelute e Okamoto (2020) apresentam o ClickPrato, um aplicativo desenvolvido por pesquisadores da USP, que é baseado em conceitos de Nutrição e Machine Learning para avaliar a qualidade da refeição de uma pessoa a partir das fotos que a pessoa tira de seu prato, a partir de um celular. Os autores desenvolveram um índice para verificar a qualidade nutricional de refeição pela sua imagem. Ao obter as fotos do prato, a refeição é qualificada através da aplicação de Deep Learning (AGGARWAL, 2018).

10 Ler mais

Show all 5181 documents...