Sie sind auf Seite 1von 5

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856

Genetic Optimized Neural Network Algorithm to Improve Classification Accuracy for Opinion Mining of M-Learning Reviews
A.Nisha Jebaseeli1, Dr.E.Kirubakaran2
1

Bharathidasan University Constituent College, Department of Computer Science, Lalgudi, Trichy, Tamilnadu, India
2

Bharat Heavy Electricals Ltd., SSTP/Systems Trichy, Tamilnadu, India

Abstract: Fast development of communication technology


leads the novel learning approach, such as Mobile Learning has been considered as an effective form of flexible learning because it is possible to break the limitation of teaching place where must take place in a classroom or located computers. Today opinions play vital role in decision making, without opinion life is like an empty vessel. Opinions are flooded in the internet leads the necessity to classify them which in turn helps the company to know its customers feedback/opinion about the product. To recognize the customers opinions through Opinion mining enrich the Quality/Quantity of the product/service. This research proposes Genetic optimized Artificial Neural Network (ANN) to classify the learners opinion about M-Learning system. The parameters learning rate and momentum values are optimized by genetic approach, which in turn enhance the accuracy of the classification process.

Keywords: M-Learning, Neural Network, Genetic Optimization, Opinion Mining, Classification Accuracy.

1. INTRODUCTION
The big boom of mobile technology reveals the greater opportunities available to the M-Learning system. Today traditional learning approach alone cannot provide effective learning methodology. So the conventional learning systems think about the modern pedagogical strategies which are mainly focused on the student centric approach. The growing popularity of mobile devices and student centric approach leads to new innovative pedagogical strategy such as the M-Learning system. Thus Mobile Learning is not just about learning using portable devices, but learning across contexts [1]. With the help of wireless technology, the internet becomes a valuable place for online learning, exchanging ideas, reviews for a product or service. Reviews in the internet could be in millions for a product or services which make it difficult to track and understand customer opinions. Opinion Mining is a relatively new and challenging field dedicated to detecting subjective content in opinionated documents, with a variety of uses in real world applications. Opinion mining determines the polarity of comments as they can be classified into positive, negative or neutral through extraction of an objects attributes and features, as commented in the reviews [2], [3]. The main Volume 2, Issue 3 May June 2013

concern of this work is to perform opinion mining tasks and enhancing the classification accuracy of the classifiers through evolutionary techniques. It is an emerging area of research to extract the subjective information in source materials by applying Natural Language Processing, Computational Linguistics and text analytics and classify the polarity of the opinion stated. Sentiment analysis in M-Learning system helps to enhance Mobile Learning system and to know users opinions of such systems. A great deal of research is going on in neural networks worldwide and its applications are expanding in various domains and disciplines such as character recognition, image compression, stock market prediction, medicine and many more. Backpropagation has been used in recognition and learning on neural networks. The relevance in optimization technique will lead to the genetic approach which has a little chance to get stuck at local minima. Genetic approaches are finding prevalent applications in solving problems requiring efficient and effective search in synthesis of neural network architectures, scheduling, numerical optimization etc., and results in solutions that are globally optimal or nearly so. Topology optimization, genetic training algorithms and control parameter optimization are the ways in which genetic algorithms are applied in ANN. In Genetic training algorithms, the learning of an ANN is formulated as a weight optimization problems such as learning rate, momentum rate, tolerance level are optimized using genetic approach [4]. Yung et al. [5] predict the stock price using a hybrid Genetic Approach (GA) combined with recurrent NNs describes a number of input variables that help the network to forecast the next day price. For the input variables, technical indicators or signals are used that were developed in deterministic trading techniques. The backpropagation algorithm is prone to get stuck in local minima and highly depends on the initial weights; a GA is used for optimizing NNs weights. When a new method was used for interpreting the weights of a trained network, it was proven that the neural networks are able to adjust their weights in accordance with the importance of the role each input data source plays during the classification [6].The use of Page 345

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856
genetic algorithm in the design of ANN improves the classification Accuracy. This paper is organized into the following sections. Section II briefly describes the materials and methods and Genetic Optimized Neural Network Algorithm, section III describes the results obtained and discusses the same. Finally the section IV concludes the paper. the research only requires that the weight changes be proportional to the derivative of the error. The larger the learning rate the larger the weight changes on each epoch, and the quicker the network learns. However, the size of the learning rate can also influence whether the network achieves a stable solution. With momentum, the direction of movement of weights in a particular weight space tends to continue in the same direction. By genetic optimizing the momentum and learning rate the convergence becomes more profound. Hence, the genetic optimized techniques to enhance accuracy are presented. The proposed Genetic Optimized Neural Network Algorithm is named as GKINN and it has two enhancements over the existing algorithm i) Implementation of top five positive, negative weights from input layer to the output layer with the help of Information Gain ii) Genetic Optimization of learning rate and momentum value. The Architecture of Genetic Optimized Neural Network is shown in Figure 1.

2. MATERIALS AND METHODS


This section illustrates the methodology followed in the proposed work. The learners opinions about M-Learning system are classified using neural genetic approach. Genetic Optimization is the evolutionary approach which aims to improve the solution to a problem by keeping the best combination of input variables. ANN is the best tool for classification and regression. The parameters of the ANN are optimized with genetic approach to enhance the classification accuracy. When the ANN is combined with genetic approach, is known as hybrid model which is performed better than the conventional ANN [7]. The Genetic Approach often works with a form of binary coding. If the problems are coded as chromosomes, the populations are initialized. Once the population size is chosen, the initial population is randomly generated. After the initialization step, each chromosome is evaluated by a fitness function. According to the value of the fitness function, the chromosome associated with fittest individuals will be reproduced more often than those associated with unfit individuals. Selection, Mating and Mutation are the three operators in genetic optimization. Selection equates to survival of the fittest. Fitness is determined by an objective function or by a subjective judgment [4]. Individuals with higher fitness survive and individuals with lower fitness die [8]. Mating is the creation of one or more offspring from the parents selected in the pairing process. The most common form of mating involves two parents that use crossover to produce two offspring. The k-point and flat crossover are among the commonly used crossover techniques. In kpoint technique, k crossover points within the range of an individual are selected randomly and the individual parts between these points are swapped by the parts of another individual. Mutation introduces random modifications [4]. Mutation consists of considering in turn each bit of a given chromosome and changing its value with a predefined low probability called the mutation. After mating process, a part of offspring are mutated. Random mutations alter a certain number of bits in a mutant offspring. The Artificial Neural Network classifiers on M-Learning corpus endorsed the improvement in the classification accuracy. As to outwit the performance of the neural network based algorithms, genetic optimization is regarded in mining the opinions. The techniques that help out in cutting across the local minima, which have sub optimal solutions, are momentum and learning rate. The improved backpropagation algorithm developed in Volume 2, Issue 3 May June 2013

Figure 1 Architecture of GKINN In the Table 1 the parameters of Genetic Optimized Neural Network are shown.

Table 1 Genetic Algorithm Parameters

Page 346

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856
The generations of momentum and learning rates for the adequate maximization of classification accuracy are genetically optimized. The sequence of steps followed in bringing forth the genetic optimization techniques are shown subsequently in Figure 2.

Classification Accuracy % Algorithms M-Learning Dataset LVQ 56.67 76.67 78.67 84.33

3. RESULTS AND DISCUSSIONS


The performance of the proposed algorithm is evaluated using Precision, Recall and F-Measure. Precision is the fraction of retrieved instances that are relevant and recall is the fraction of relevant instances that are retrieved. F-Measure is the Harmonic mean of Precision and Recall. Precision and recall are the inverse relationship, where it is possible to increase one at the cost of reducing the other [11]. The proposed Genetic Optimized Neural Network algorithm is operated with M-Learning system reviews as dataset. Learners opinions about free M-Learning system available are considered. The preprocessed dataset using proposed preprocessing algorithm is applied in Genetic Optimized Neural Network algorithm for classification. The classification accuracy obtained from the experiment reaches up to 84.33%. The experiment with the acquired dataset has revealed that the increase of classification accuracy. The same algorithm is also commendable when the existing datasets are used.

Elman FFNN GKINN

Classification Accuracy in %

100 80 60 40 20

83.33 84.33 76.67 78.67 56.67

Figure 3 Classification Accuracy of GKINN

The summary of results and classification accuracy obtained is tabulated in Table 2 and it is graphically shown in Figure 3. The obtained classification accuracies from the classifiers are validated using Precision, Recall and F-Measure. Precision is the probability that the retrieved document is relevant and Recall is the probability that a relevant document is retrieved in a search. F-Measure is the harmonic mean of Precision and Recall. The Validated measures are shown in Table 3 for the FFNN, KINN and the proposed GKINN classifiers
Table 3 Precision, Recall and F-measure of M-Learning Dataset

Precision Algorithms

Recall

F-Measure

M-Learning Dataset FFNN KINN Figure 2 Genetic Optimization of Momentum and Learning Rate Table 2 Comparing Classification Accuracy of GKINN with various Algorithms 0.667 0.837 0.8 0.826 0.727 0.831

GKINN

0.848

0.834

0.839

The measures of validation are represented graphically in Figure 4, which shows an improvement in the accuracy rate. It is realized that the precision, recall and F-Measure

Volume 2, Issue 3 May June 2013

Page 347

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856
of the proposed algorithm better than the existing algorithm. These techniques have impact on the accuracy when genetically optimized. With the fitness value resolving the selection of the randomly generated momentum, the optimized momentum and learning rate values are determined. The results were compared with the existing algorithms and the quality of classifications is measured with Precision, Recall and F-measure. The results shows that the genetic optimized ANN gives better performance when compare to other classifiers.

References
[1] Walker, K, Mapping the landscape of mobile learning. In Big Issues in Mobile Learning. Sharples, M., (Ed), A Report of a workshop by the Kaleidoscope Network of Excellence Mobile Learning Initiative, pp. 3-4, 2007. [2] Dave, D., Lawrence, A., Pennock, D, Mining the Peanut Gallery: Opinion Extraction and Semantic Classification of Product Reviews, Proceedings of International World Wide Web Conference (WWW03), pp. 519-528, 2003. [3] Turney, P., Thumbs Up or Thumbs Down? Semantic Orientation Applied to Unsupervised Classification of Reviews, Proceedings of the 40th Annual Meeting on Association for Computational Linguistics, pp. 417424, 2002. [4] S.N. Sivanandam, S.N. Deepa, Principles of Soft computing, Wiley India(P) Ltd., 2007. [5] Yung-Keun Kwon, Byung-Ro Moon, A Hybrid Neurogenetic Approach for Stock Forecasting, IEEE Transactions on Neural Networks, Volume 18, Issue 3, pp. 851-864, 2007. [6] Weiyang Zhou, Verification of the Nonparametric Characteristics of Backpropagation Neural Networks for Image Classification, IEEE Transactions on Geoscience and Remote Sensing, Volume 37, Number 2, 1999. [7] Kyoung-jae Kima, Ingoo Hanb, Application of a hybrid genetic algorithm and neural network approach in activity-based costing, Expert Systems with Applications, Volume 24, pp. 7377, 2003. [8] Max Mignotte, Christophe Collet, Patrick Perez, Patrick Bouthemy, Hybrid Genetic Optimization and Statistical Model-Based Approach for the Classification of Shadow Shapes in Sonar Imagery, IEEE Transactions On Pattern Analysis and Machine Intelligence, Volume 22, Number 2, 2000. [9] Chih-Ming Chen, Shih-Hsun Hsu, Personalized Intelligent Mobile Learning System for Supporting Effective English Learning, International Forum of Educational Technology & Society, Volume 11, Issue 3, pp. 153-180, 2008. Page 348

Figure 4 Precision, Recall and F-measure of M-Learning Dataset Thus the proposed Genetic Optimized Neural Network algorithm is validated against the existing ANN classifier with Precision, Recall and F-Measures. The Classification accuracy range 84.33% is obtained.

4. CONCLUSION
Fast development of communication technology leads the novel learning approach, such as Mobile Learning has been considered as an effective form of flexible learning because it is possible to break the limitation of teaching place where must take place in a classroom or located computers. In the meanwhile, Mobile Learning is also helpful to perform learning activities utilizing learners spare time at any time from any place with wireless networks [9]. For popular products, the reviews found in the internet could be in thousands which makes it difficult to track and understand customers opinions. Similarly, the manufacturer of a product or service provider finds it difficult to analyze the market pulse. The development of communication technology has led to easy access of information through the internet. Nowadays, the use of mobile devices is increasing rapidly which in turn has popularized the pedagogical methods such as learning through mobile devices [10]. Several mobile learning systems are available and also the users opinions about these systems are aired in the social blogs or review Websites. Opinions had been classified by the neural network based algorithms and in this work the classification accuracy is further enhanced with genetic optimization. This research adopts Momentum and Learning rates as to enhance classification accuracy. Volume 2, Issue 3 May June 2013

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856
[10] Sharples, M., Lonsdale P. Meek J., Rudman P. D., and Vavoula G., An Evaluation of MyArtSpace: a Mobile Learning Service for School Museum Trips, Proceedings of the 6th Annual Conference on Mobile Learning, pp. 238-244, 2007. [11] Powers, D.M.W, Evaluation: From Precision, Recall and F-Measure To Roc, Informedness, Markedness & Correlation, Journal of Machine Learning Technologies, Volume 2, Issue 1, pp-37-63, 2011.

AUTHOR
A. Nisha Jebaseeli received her M.Tech degree in Bharathidasan University, Trichy, India in 2008. Now she is working as a Assistant Professor with Department of Computer Science, Bharathidasan University Constituent College, Lalgudi, Trichy, India and also she is pursuing PhD (Computer Science) in Bharathidasan University. Dr. E. Kirubakaran obtained his B.E (Honours.) degree in Mechanical Engineering from Regional Engineering College Trichy in the year 1978 and joined in Bharat Heavy Electricals Ltd. Tiruchirappalli in 1979. He obtained M.E. in Computer Science from Regional Engineering College in 1984. In 1999, he obtained his Ph.D. degree in Computer Science from Bharathidasan University through Regional Engineering College. In 2005, he got his M.B.A. degree from IGNOU. Presently he is Additional General Manager, SSTP/Systems, at BHEL, Trichy For more than 31 years he has been working in designing, Developing and maintenance of Software Systems at BHEL, Trichy.

Volume 2, Issue 3 May June 2013

Page 349

Das könnte Ihnen auch gefallen