An Anomaly Detection Approach to Detect Unexpected Faults in Recordings from Test Drives

In the automotive industry test drives are being conducted during the development of new vehicle models or as a part of quality assurance of series-production vehicles. The communication on the in-vehicle network, data from external sensors, or internal data from the electronic control units is recorded by automotive data loggers during the test drives. The recordings are used for fault analysis. Since the resulting data volume is tremendous, manually analysing each recording in great detail is not feasible. This paper proposes to use machine learning to support domainexperts by preventing them from contemplating irrelevant data and rather pointing them to the relevant parts in the recordings. The underlying idea is to learn the normal behaviour from available recordings, i.e. a training set, and then to autonomously detect unexpected deviations and report them as anomalies. The one-class support vector machine “support vector data description” is utilised to calculate distances of feature vectors. SVDDSUBSEQ is proposed as a novel approach, allowing to classify subsequences in multivariate time series data. The approach allows to detect unexpected faults without modelling effort as is shown with experimental results on recordings from test drives.

The Influence of Preprocessing Parameters on Text Categorization

Text categorization (the assignment of texts in natural language into predefined categories) is an important and extensively studied problem in Machine Learning. Currently, popular techniques developed to deal with this task include many preprocessing and learning algorithms, many of which in turn require tuning nontrivial internal parameters. Although partial studies are available, many authors fail to report values of the parameters they use in their experiments, or reasons why these values were used instead of others. The goal of this work then is to create a more thorough comparison of preprocessing parameters and their mutual influence, and report interesting observations and results.

Weka Based Desktop Data Mining as Web Service

Data mining is the process of sifting through large volumes of data, analyzing data from different perspectives and summarizing it into useful information. One of the widely used desktop applications for data mining is the Weka tool which is nothing but a collection of machine learning algorithms implemented in Java and open sourced under the General Public License (GPL). A web service is a software system designed to support interoperable machine to machine interaction over a network using SOAP messages. Unlike a desktop application, a web service is easy to upgrade, deliver and access and does not occupy any memory on the system. Keeping in mind the advantages of a web service over a desktop application, in this paper we are demonstrating how this Java based desktop data mining application can be implemented as a web service to support data mining across the internet.

Ranking - Convex Risk Minimization

The problem of ranking (rank regression) has become popular in the machine learning community. This theory relates to problems, in which one has to predict (guess) the order between objects on the basis of vectors describing their observed features. In many ranking algorithms a convex loss function is used instead of the 0-1 loss. It makes these procedures computationally efficient. Hence, convex risk minimizers and their statistical properties are investigated in this paper. Fast rates of convergence are obtained under conditions, that look similarly to the ones from the classification theory. Methods used in this paper come from the theory of U-processes as well as empirical processes.

A Cognitive Robot Collaborative Reinforcement Learning Algorithm

A cognitive collaborative reinforcement learning algorithm (CCRL) that incorporates an advisor into the learning process is developed to improve supervised learning. An autonomous learner is enabled with a self awareness cognitive skill to decide when to solicit instructions from the advisor. The learner can also assess the value of advice, and accept or reject it. The method is evaluated for robotic motion planning using simulation. Tests are conducted for advisors with skill levels from expert to novice. The CCRL algorithm and a combined method integrating its logic with Clouse-s Introspection Approach, outperformed a base-line fully autonomous learner, and demonstrated robust performance when dealing with various advisor skill levels, learning to accept advice received from an expert, while rejecting that of less skilled collaborators. Although the CCRL algorithm is based on RL, it fits other machine learning methods, since advisor-s actions are only added to the outer layer.

On Speeding Up Support Vector Machines: Proximity Graphs Versus Random Sampling for Pre-Selection Condensation

Support vector machines (SVMs) are considered to be the best machine learning algorithms for minimizing the predictive probability of misclassification. However, their drawback is that for large data sets the computation of the optimal decision boundary is a time consuming function of the size of the training set. Hence several methods have been proposed to speed up the SVM algorithm. Here three methods used to speed up the computation of the SVM classifiers are compared experimentally using a musical genre classification problem. The simplest method pre-selects a random sample of the data before the application of the SVM algorithm. Two additional methods use proximity graphs to pre-select data that are near the decision boundary. One uses k-Nearest Neighbor graphs and the other Relative Neighborhood Graphs to accomplish the task.

Machine Learning Techniques for Short-Term Rain Forecasting System in the Northeastern Part of Thailand

This paper presents the methodology from machine learning approaches for short-term rain forecasting system. Decision Tree, Artificial Neural Network (ANN), and Support Vector Machine (SVM) were applied to develop classification and prediction models for rainfall forecasts. The goals of this presentation are to demonstrate (1) how feature selection can be used to identify the relationships between rainfall occurrences and other weather conditions and (2) what models can be developed and deployed for predicting the accurate rainfall estimates to support the decisions to launch the cloud seeding operations in the northeastern part of Thailand. Datasets collected during 2004-2006 from the Chalermprakiat Royal Rain Making Research Center at Hua Hin, Prachuap Khiri khan, the Chalermprakiat Royal Rain Making Research Center at Pimai, Nakhon Ratchasima and Thai Meteorological Department (TMD). A total of 179 records with 57 features was merged and matched by unique date. There are three main parts in this work. Firstly, a decision tree induction algorithm (C4.5) was used to classify the rain status into either rain or no-rain. The overall accuracy of classification tree achieves 94.41% with the five-fold cross validation. The C4.5 algorithm was also used to classify the rain amount into three classes as no-rain (0-0.1 mm.), few-rain (0.1- 10 mm.), and moderate-rain (>10 mm.) and the overall accuracy of classification tree achieves 62.57%. Secondly, an ANN was applied to predict the rainfall amount and the root mean square error (RMSE) were used to measure the training and testing errors of the ANN. It is found that the ANN yields a lower RMSE at 0.171 for daily rainfall estimates, when compared to next-day and next-2-day estimation. Thirdly, the ANN and SVM techniques were also used to classify the rain amount into three classes as no-rain, few-rain, and moderate-rain as above. The results achieved in 68.15% and 69.10% of overall accuracy of same-day prediction for the ANN and SVM models, respectively. The obtained results illustrated the comparison of the predictive power of different methods for rainfall estimation.

A Hybrid Feature Selection by Resampling, Chi squared and Consistency Evaluation Techniques

In this paper a combined feature selection method is proposed which takes advantages of sample domain filtering, resampling and feature subset evaluation methods to reduce dimensions of huge datasets and select reliable features. This method utilizes both feature space and sample domain to improve the process of feature selection and uses a combination of Chi squared with Consistency attribute evaluation methods to seek reliable features. This method consists of two phases. The first phase filters and resamples the sample domain and the second phase adopts a hybrid procedure to find the optimal feature space by applying Chi squared, Consistency subset evaluation methods and genetic search. Experiments on various sized datasets from UCI Repository of Machine Learning databases show that the performance of five classifiers (Naïve Bayes, Logistic, Multilayer Perceptron, Best First Decision Tree and JRIP) improves simultaneously and the classification error for these classifiers decreases considerably. The experiments also show that this method outperforms other feature selection methods.

Development of Genetic-based Machine Learning for Network Intrusion Detection (GBML-NID)

Society has grown to rely on Internet services, and the number of Internet users increases every day. As more and more users become connected to the network, the window of opportunity for malicious users to do their damage becomes very great and lucrative. The objective of this paper is to incorporate different techniques into classier system to detect and classify intrusion from normal network packet. Among several techniques, Steady State Genetic-based Machine Leaning Algorithm (SSGBML) will be used to detect intrusions. Where Steady State Genetic Algorithm (SSGA), Simple Genetic Algorithm (SGA), Modified Genetic Algorithm and Zeroth Level Classifier system are investigated in this research. SSGA is used as a discovery mechanism instead of SGA. SGA replaces all old rules with new produced rule preventing old good rules from participating in the next rule generation. Zeroth Level Classifier System is used to play the role of detector by matching incoming environment message with classifiers to determine whether the current message is normal or intrusion and receiving feedback from environment. Finally, in order to attain the best results, Modified SSGA will enhance our discovery engine by using Fuzzy Logic to optimize crossover and mutation probability. The experiments and evaluations of the proposed method were performed with the KDD 99 intrusion detection dataset.

Traffic Flow Prediction using Adaboost Algorithm with Random Forests as a Weak Learner

Traffic Management and Information Systems, which rely on a system of sensors, aim to describe in real-time traffic in urban areas using a set of parameters and estimating them. Though the state of the art focuses on data analysis, little is done in the sense of prediction. In this paper, we describe a machine learning system for traffic flow management and control for a prediction of traffic flow problem. This new algorithm is obtained by combining Random Forests algorithm into Adaboost algorithm as a weak learner. We show that our algorithm performs relatively well on real data, and enables, according to the Traffic Flow Evaluation model, to estimate and predict whether there is congestion or not at a given time on road intersections.

Validation and Selection between Machine Learning Technique and Traditional Methods to Reduce Bullwhip Effects: a Data Mining Approach

The aim of this paper is to present a methodology in three steps to forecast supply chain demand. In first step, various data mining techniques are applied in order to prepare data for entering into forecasting models. In second step, the modeling step, an artificial neural network and support vector machine is presented after defining Mean Absolute Percentage Error index for measuring error. The structure of artificial neural network is selected based on previous researchers' results and in this article the accuracy of network is increased by using sensitivity analysis. The best forecast for classical forecasting methods (Moving Average, Exponential Smoothing, and Exponential Smoothing with Trend) is resulted based on prepared data and this forecast is compared with result of support vector machine and proposed artificial neural network. The results show that artificial neural network can forecast more precisely in comparison with other methods. Finally, forecasting methods' stability is analyzed by using raw data and even the effectiveness of clustering analysis is measured.

Combining Bagging and Boosting

Bagging and boosting are among the most popular resampling ensemble methods that generate and combine a diversity of classifiers using the same learning algorithm for the base-classifiers. Boosting algorithms are considered stronger than bagging on noisefree data. However, there are strong empirical indications that bagging is much more robust than boosting in noisy settings. For this reason, in this work we built an ensemble using a voting methodology of bagging and boosting ensembles with 10 subclassifiers in each one. We performed a comparison with simple bagging and boosting ensembles with 25 sub-classifiers, as well as other well known combining methods, on standard benchmark datasets and the proposed technique was the most accurate.

Dataset Analysis Using Membership-Deviation Graph

Classification is one of the primary themes in computational biology. The accuracy of classification strongly depends on quality of a dataset, and we need some method to evaluate this quality. In this paper, we propose a new graphical analysis method using 'Membership-Deviation Graph (MDG)' for analyzing quality of a dataset. MDG represents degree of membership and deviations for instances of a class in the dataset. The result of MDG analysis is used for understanding specific feature and for selecting best feature for classification.

Learning Process Enhancement for Robot Behaviors

Designing a simulated system and training it to optimize its tasks in simulated environment helps the designers to avoid problems that may appear when designing the system directly in real world. These problems are: time consuming, high cost, high errors percentage and low efficiency and accuracy of the system. The proposed system will investigate and improve the efficiency and accuracy of a simulated robot to choose correct behavior to perform its task. In this paper, machine learning, which uses genetic algorithm, is adopted. This type of machine learning is called genetic-based machine learning in which a distributed classifier system is used to improve the efficiency and accuracy of the robot. Consequently, it helps the robot to achieve optimal action.

Critical Analysis of Decision Making Experience with a Machine Learning Approach in Playing Ayo Game

The major goal in defining and examining game scenarios is to find good strategies as solutions to the game. A plausible solution is a recommendation to the players on how to play the game, which is represented as strategies guided by the various choices available to the players. These choices invariably compel the players (decision makers) to execute an action following some conscious tactics. In this paper, we proposed a refinement-based heuristic as a machine learning technique for human-like decision making in playing Ayo game. The result showed that our machine learning technique is more adaptable and more responsive in making decision than human intelligence. The technique has the advantage that a search is astutely conducted in a shallow horizon game tree. Our simulation was tested against Awale shareware and an appealing result was obtained.

Meta Random Forests

Leo Breimans Random Forests (RF) is a recent development in tree based classifiers and quickly proven to be one of the most important algorithms in the machine learning literature. It has shown robust and improved results of classifications on standard data sets. Ensemble learning algorithms such as AdaBoost and Bagging have been in active research and shown improvements in classification results for several benchmarking data sets with mainly decision trees as their base classifiers. In this paper we experiment to apply these Meta learning techniques to the random forests. We experiment the working of the ensembles of random forests on the standard data sets available in UCI data sets. We compare the original random forest algorithm with their ensemble counterparts and discuss the results.

A Bayesian Kernel for the Prediction of Protein- Protein Interactions

Understanding proteins functions is a major goal in the post-genomic era. Proteins usually work in context of other proteins and rarely function alone. Therefore, it is highly relevant to study the interaction partners of a protein in order to understand its function. Machine learning techniques have been widely applied to predict protein-protein interactions. Kernel functions play an important role for a successful machine learning technique. Choosing the appropriate kernel function can lead to a better accuracy in a binary classifier such as the support vector machines. In this paper, we describe a Bayesian kernel for the support vector machine to predict protein-protein interactions. The use of Bayesian kernel can improve the classifier performance by incorporating the probability characteristic of the available experimental protein-protein interactions data that were compiled from different sources. In addition, the probabilistic output from the Bayesian kernel can assist biologists to conduct more research on the highly predicted interactions. The results show that the accuracy of the classifier has been improved using the Bayesian kernel compared to the standard SVM kernels. These results imply that protein-protein interaction can be predicted using Bayesian kernel with better accuracy compared to the standard SVM kernels.

Corporate Credit Rating using Multiclass Classification Models with order Information

Corporate credit rating prediction using statistical and artificial intelligence (AI) techniques has been one of the attractive research topics in the literature. In recent years, multiclass classification models such as artificial neural network (ANN) or multiclass support vector machine (MSVM) have become a very appealing machine learning approaches due to their good performance. However, most of them have only focused on classifying samples into nominal categories, thus the unique characteristic of the credit rating - ordinality - has been seldom considered in their approaches. This study proposes new types of ANN and MSVM classifiers, which are named OMANN and OMSVM respectively. OMANN and OMSVM are designed to extend binary ANN or SVM classifiers by applying ordinal pairwise partitioning (OPP) strategy. These models can handle ordinal multiple classes efficiently and effectively. To validate the usefulness of these two models, we applied them to the real-world bond rating case. We compared the results of our models to those of conventional approaches. The experimental results showed that our proposed models improve classification accuracy in comparison to typical multiclass classification techniques with the reduced computation resource.

Software Maintenance Severity Prediction with Soft Computing Approach

As the majority of faults are found in a few of its modules so there is a need to investigate the modules that are affected severely as compared to other modules and proper maintenance need to be done on time especially for the critical applications. In this paper, we have explored the different predictor models to NASA-s public domain defect dataset coded in Perl programming language. Different machine learning algorithms belonging to the different learner categories of the WEKA project including Mamdani Based Fuzzy Inference System and Neuro-fuzzy based system have been evaluated for the modeling of maintenance severity or impact of fault severity. The results are recorded in terms of Accuracy, Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The results show that Neuro-fuzzy based model provides relatively better prediction accuracy as compared to other models and hence, can be used for the maintenance severity prediction of the software.

Pruning Method of Belief Decision Trees

The belief decision tree (BDT) approach is a decision tree in an uncertain environment where the uncertainty is represented through the Transferable Belief Model (TBM), one interpretation of the belief function theory. The uncertainty can appear either in the actual class of training objects or attribute values of objects to classify. In this paper, we develop a post-pruning method of belief decision trees in order to reduce size and improve classification accuracy on unseen cases. The pruning of decision tree has a considerable intention in the areas of machine learning.