Integrated Cosine And Tuned Cosine Similarity Measure To Alleviate Data Sparsity Issues For Personalized Recommendation

E-commerce is an exemplary representation for online commercial transactions which permits business applications to deal with different organizations. Many organizations rely on these websites to show rapid improvement in online digital transaction by amplifying an accurate recommendation list to the end user. One of the utmost successes dealing in this e-commerce era is said to be recommender system. Whose functionality is to build a customer product relationship in an efficient way. The most appropriate product or item is predicted in advanced and thus it is recommended to the end user. The most famous machine learning algorithm taken away by the recommender system is collaborative filtering (CF) method. It hunts for a "like-Minder" people whose preference in buying a similar item in advance. The customary similarity measures are Pearson correlation coefficient (PCC), cosine and Jaccard. These measures are not much effective in prediction part. So that a new similarity measure named as Tuned Cosine (TCOS) was proposed. Research objective involves: to show the increase in prediction accuracy for the users unrated item i.e. data sparsity using user-item rating matrix facto rization method. Finding an enhanced solution for a data sparsity issues which is the primary issue needs to resolved effectively in collaborative filtering method by proposing a new heuristic similarity measure. Predicting user's unrated items with high accuracy rate gradually increases the recommendation performance. Compare conventional algorithm method with the proposed TCOS algorithm by calculating MAE and RMSE error ratio. Thus the new TCOS similarity measure shows the superiority in recommendation performance by showing satisfactory accuracy in prediction part and reduces the error ratio than the predefined one. The proposed TCOS method is effective in handling Big Data too.