A novel approach in analysing tweets beyond its polarity

Abstract

Social networking sites and services are phenomenon among young newlinepeople all across the world. Started as a mean to share interests and activities newlinethey are now maturing to a point of redefining our current human civilization. newlineThis proposes a research interest that how these large data sets can be mined newlinecollectively to extract patterns. The purpose of uncovering these hidden patterns newlineis to Analyze, Observe and Understand the human behavior newlineAs the first contribution, a Rule Generation-and-Clustering-centered newlineUncovering Hidden Patterns in SM (RCUHP-SM) methodology is proposed to newlineeffectually uncover the hidden patterns. Primarily, the customer review data set newlineis proffered as input and it is pre-processed by eradicating the unwanted and newlineirrelevant attributes. Subsequently, the descriptive sentences are extorted as of newlinethe pre-processed data and the tagged words are counted to evaluate its score. It newlineis grounded on the positive, negative and neutral reviews of the user of each newlineproduct. Here, Word Net dictionaries are used for finding polarities as it shows newlinehigh accuracy than its contenders. Then, a set of rules from R1 to R27 is framed newlineto predict the category of review. Consequently, the threshold value is calculated newlineto create the cluster groups into least similar, moderately similar and most newlinesimilar. Then, it will be labeled as C1 to C6 based on its category. In the analysis newlinephase, the features are extorted as of the product description and its newlinecorresponding score is computed. Grounded on the score, the features are sorted newlineand analyzed for the recommendation. In this work, the novelty is proffered in newlinerule generation, threshold-based cluster formation, similarity computation, and newlineanalysis stages. In experiments, the performance rendered by the proposed newlineuncovering hidden pattern system is assessed and contrasted in respect of Root newlineMean Squared Error (RMSE), Absolute Precision (MAP), and Mean Absolute newlineError (MAE) measures. newline