ICTACT Journal on Soft Computing

R. Abitha ¹, S. Mary Vennila ¹

Affiliations
1 Department of Computer Science, Presidency College, IN

Abstract

Association rule mining (ARM) is a common and most preferable research method for bringing out the fascinating relations between the variables provided in any data set. It brings out the knowledge by satisfying the user defined values and criteria measures specified by the researcher. Frequent item set generation is a well-known method carried out by many researchers to retrieve interesting correlations among the variables that helps in decision making. The accuracy of the brought out rules by ARM is good enough to provide a conclusion on research studies. This can be improved by incorporating optimization like heuristic search techniques. In this paper cultural algorithm is used to improve the performance of rule mining by optimization which is required for categorizing the risk level of ASD individuals. Optimization is utilized in health care domain for generating optimized rules to analyze the frequently combined attributes among the patient’s data. It gracefully improves the result finding process which will be tranquil to conclude the decision. The Cultural algorithm fit in to the larger course of evolutionary algorithms that is inspired by natural evolution. In this paper multi objective optimization technique is proposed by incorporating ARM and cultural algorithm by considering different objectives namely support, confidence, lift and completeness of the rule to find the positive association of Autism Spectrum Disorder (ASD) screening data features with positive class. The result of this research depicts the positive association with improved performance along with reduced number of rules.

Keywords

Association Rule Mining, ASD, Apriori Algorithm, Rule Generation, Cultural Algorithm, Risk Level, Optimization.

Full Text

     

K. Raviya ¹, S. Mary Vennila ¹

Affiliations
1 PG and Research Department of Computer Science, Presidency College, IN

Abstract

With the growth of e-commerce reporting, online customer reviews have evolved rapidly, voicing the sentiment or opinion of customers about goods. The analysis of belief could provide useful data for us. Sentiment analysis on social media like Twitter or Facebook, is now the comprehensive way of understanding about the views of customers and has extensive variety of applications. In the context of NLP, automated text classification can be a fundamental activity and it can help people to pick essential information from vast text resources. Sentiment analysis may be a computational technique that plays a key role in automating the retrieval of subjective knowledge, i.e. customer’s sentiment from online text reviews or opinion from social network like Twitter and Facebook. Lexicon-based and machine learning-based methods are two main approaches widely used in sentiment analysis activities. In machine learning based framework, Sentiment analysis is a text recognition task. The outcome depends not only from the soundness of the algorithm for machine learning, but also with the appropriate features. In recent years, the most recent technological advancements, like deep-learning techniques, have resolved a number of standard challenges caused by the lack of lexical tools in the region. It has been exhibited that deep-learning models are auspicious and potential tool to NLP challenges. In this work, the fusion of deep CNN with SVM will automatically detect and extract subjective sentence-level features to perform sentiment analysis of online product review dataset with highest accuracy and less computation time.

Keywords

Sentiment Analysis, Deep Learning, Convolutional Neural Network, Bigdata.

Full Text