Hybrid Measuring the Similarity Value Based on Genetic Algorithm for Improving Prediction in A Collaborative Filtering Recommendation System.

Abstract

In recent years, the Recommendation System (RS) has a wide range of applications in several fields, like Education, Economics, Scientific Researches and other related fields. The Personalized Recommendation is effective in increasing RS's accuracy, based on the user interface, preferences and constraints seek to predict the most suitable product or services. Collaborative Filtering (CF) is one of the primary applications that researchers use for the prediction of the accuracy rating and recommendation of objects. Various experts in the field are using methods like Nearest Neighbors (NN) based on the measures of similarity. However, similarity measures use only the co-rated item ratings while calculating the similarity between a pair of users or items. The two standard methods used to measure similarities are Cosine Similarity (CS) and Person Correlation Similarity (PCS). However, both are having drawbacks, and the present piece of the investigation will approach through the optimized Genetic Algorithms (GA) to improve the forecast accuracy of RS using the merge output of CS with PCS based on GA methods. The results show GA's superiority and its ability to achieve more correct predictions than CS and PCS.