Binoy B. Nair

Application of Support Vector Machine (SVM) and Proximal Support Vector Machine (PSVM) for fault classification of monoblock centrifugal pump

Monoblock centrifugal pumps are widely used in a variety of applications. Defects and malfunctions (faults) of these pumps result in significant economic loss. Therefore, the pumps must be under constant monitoring. When a possible fault is detected, diagnosis is carried out to pinpoint it. In many applications, the role of monoblock centrifugal pumps is critical and condition monitoring is essential. Vibration-based condition monitoring and analysis using the machine-learning approach is gaining momentum. In particular, Artificial Neural Networks (ANNs), fuzzy logic and roughsets have been employed for condition monitoring and fault diagnosis. While it is difficult to train the neural network-based fault classifier, the classification accuracy in case of fuzzy logic- and roughest-based fault classifiers is not very high. This paper presents the use of Support Vector Machines (SVMs) and Proximal Support Vector Machines (PSVMs) for classifying faults using statistical features extracted from vibration signals under good and faulty conditions of a monoblock centrifugal pump. The Decision Tree (DT) algorithm is used to select prime features. These features are fed as inputs for training and testing SVMs and PSVMs and their fault classification accuracy is compared. The results are found to be better than neural network-, fuzzy- and roughest-based methods.

Automatic rule learning using roughset for fuzzy classifier in fault categorization of mono-block centrifugal pump

Rule learning based approach to fault detection and diagnosis is becoming very popular, mainly due to their high accuracy when compared to older statistical methods. Fault detection and diagnosis of various mechanical components of centrifugal pump is essential to increase the productivity and reduce the breakdowns. This paper presents the use of rough sets to generate the rules from statistical features extracted from vibration signals under good and faulty conditions of a centrifugal pump. A fuzzy inference system (FIS) is built using rough set rules and tested using test data. The effect of different types of membership functions on the FIS performance is also presented. Finally, the performance of this classifier is compared to that of a fuzzy-antminer classifier and to multi-layer perceptron (MLP) based classifiers.

A GA-artificial neural network hybrid system for financial time series forecasting

Accurate prediction of financial time series, such as those generated by stock markets, is a highly challenging task due to the highly nonlinear nature of such series. A novel method of predicting the next day’s closing value of a stock market is proposed and empirically validated in the present study. The system uses an adaptive artificial neural network based system to predict the next day’s closing value of a stock market index. The proposed system adapts itself to the changing market dynamics with the help of genetic algorithm which tunes the parameters of the neural network at the end of each trading session so that best possible accuracy is obtained. The effectiveness of the proposed system is established by testing on five international stock indices using ten different performance measures.

Use of histogram features for decision tree-based fault diagnosis of monoblock centrifugal pump

Monoblock centrifugal pumps are a crucial part of many industrial plants. Early detection of faults in pumps can increase their reliability, reduce energy consumption, service and maintenance costs, and increase their life-cycle and safety, thus resulting in a significant reduction in life-time costs. It is clear that the fault diagnosis and condition monitoring of pumps are important issues that cannot be ignored. Machine learning-based approach to fault detection and diagnosis is becoming very popular, mainly due to their high accuracy when compared to older statistical methods. There are set of related activities involved in machine learning approach namely, data acquisition from the monoblock centrifugal pump, feature extraction from the acquired data, feature selection, and finally feature classification. This paper presents the use of C4.5 decision tree algorithm for fault diagnosis through histogram feature extracted from vibration signals of good and faulty conditions of monoblock centrifugal pump. The performance of the proposed system is compared to that of a Naive Bayes-based system to validate the superiority of the proposed system.

Decision support system using artificial immune recognition system for fault classification of centrifugal pump

Centrifugal pumps are a crucial part of many industrial plants. Early detection of faults in pumps can increase their reliability, reduce energy consumption, service and maintenance costs, and increase their life-cycle and safety, thus resulting in a significant reduction in life-time costs. Vibration analysis is a very popular tool for condition monitoring of machinery like pumps, turbines and compressors. The proposed method is based on a novel immune inspired supervised learning algorithm which is known as artificial immune recognition system (AIRS). This paper compares the fault classification efficiency of AIRS with hybrid systems such as principle component analysis (PCA)-Naive Bayes and PCA-Bayes Net. The robustness of the proposed method is examined using its classification accuracy and kappa statistics. It is observed that the AIRS-based system outperforms the other two methods considered in the present study.

Predicting stock market trends using hybrid ant-colony-based data mining algorithms: an empirical validation on the Bombay Stock Exchange

Ant Colony Optimisation (ACO) algorithms use simple mutually cooperating agents (ants) to produce a robust and adaptive search system, which can be used for knowledge discovery. In this paper, a Support Vector Machine (SVM)-cAnt-Miner-based system for predicting the next-days trend in stock markets is proposed. The trend predicted by the proposed system is then used to identify the appropriate time to buy and sell securities. Performance of the proposed system is evaluated against SVM-Ant-Miner, SVM-Ant-Miner2, Naive-Bayes and an Artificial Neural Network (ANN)-based trend prediction system. The results indicate that the proposed system outperforms all the other techniques considered.

Artificial intelligence applications in financial forecasting --a survey and some empirical results

Financial forecasting is an area of research which has been attracting a lot of attention recently from practitioners in the field of artificial intelligence. Apart from the economic benefits of accurate financial prediction, the inherent nonlinearities in financial data make the task of analyzing and forecasting an extremely challenging task. This paper presents a survey of more than 100 articles published over two centuries from 1933 up to 2013 in an attempt to identify the developments and trends in the field of financial forecasting with focus on application of artificial intelligence for the purpose. The findings from the survey indicate that artificial intelligence and signal processing based techniques are more efficient when compared to traditional financial forecasting techniques and these techniques appear well suited for the task of financial forecasting. Some of the issues that need addressing are discussed in brief. A novel technique for selection of the input dataset size for ensuring best possible forecast accuracy is also presented. The results confirm the effectiveness of the proposed technique in improving the accuracy of forecasts.

Clustering stock price time series data to generate stock trading recommendations: An empirical study

Abstract Predicting the stock market is considered to be a very difficult task due to its non-linear and dynamic nature. Our proposed system is designed in such a way that even a layman can use it. It reduces the burden on the user. The users job is to give only the recent closing prices of a stock as input and the proposed Recommender system will instruct him when to buy and when to sell if it is profitable or not to buy share in case if it is not profitable to do trading. Using soft computing based techniques is considered to be more suitable for predicting trends in stock market where the data is chaotic and large in number. The soft computing based systems are capable of extracting relevant information from large sets of data by discovering hidden patterns in the data. Here regression trees are used for dimensionality reduction and clustering is done with the help of Self Organizing Maps (SOM). The proposed system is designed to assist stock market investors identify possible profit-making opportunities and also help in developing a better understanding on how to extract the relevant information from stock price data.

A Stock Trading Recommender System Based on Temporal Association Rule Mining

Recommender systems capable of discovering patterns in stock price movements and generating stock recommendations based on the patterns thus discovered can significantly supplement the decision-making process of a stock trader. Such recommender systems are of great significance to a layperson who wishes to profit by stock trading even while not possessing the skill or expertise of a seasoned trader. A genetic algorithm optimized Symbolic Aggregate approXimation (SAX)–Apriori based stock trading recommender system, which can mine temporal association rules from the stock price data set to generate stock trading recommendations, is presented in this article. The proposed system is validated on 12 different data sets. The results indicate that the proposed system significantly outperforms the passive buy-and-hold strategy, offering scope for a layperson to successfully invest in capital markets.

An intelligent recommender system for stock trading

Abstract. Generating consistent profits from stock markets is considered to be a challenging task, especially due to the nonlinear nature of the stock price movements. Traders need to have a deep understanding of the market behavior patterns in order to trade successfully. In this study, a GA optimized technical indicator decision tree-SVM based intelligent recommender system is proposed, which can learn patterns from the stock price movements and then recommend appropriate one-day-ahead trading strategy. The recommender system takes the task of identifying stock price patterns on itself, allowing even a lay-user, who is not well versed in stock market behavior, to trade profitably on a consistent basis. The efficacy of the proposed system is validated on four different stocks belonging to two different stock markets (India and UK) over three different time frames for each stock. Performance of the proposed system is validated using fifteen different measures. Performance is compared with traditional technical indicator based trading and the traditional buy and hold strategy. Results indicate that the proposed system is capable of generating profits for all the stocks in both the stock markets considered.