Improved Collaborative Recommender systems for Healthcare application

Recommender System (RS) is an active research study that has been widely investigated by several research communities in recent years due to the rapid growth of Internet services. Healthcare application systems aim to present recommendations from reviews presented through online user reviews posted by several other patients. The study presented here aims to analyze Online Collaborative Filtering (ONCF), which adopts a multi-layer neural network architecture for Collaborative Filtering (ONCF) combined with Collaborative Filtering (CF). The purpose of the study is to identify appropriate methods that discover significant relationships or similarities between the patients who have interrelated similarities which would be useful for the appropriate decision-making process. To provide improvements, methods involving ONCF, and Extended Recurrent Convolutional Neural Networks (ERCNN) were analyzed in the study. From the drawbacks analyzed from ONCF and ERCNN, an improved combinational hybrid approach combines both approaches which are termed in the paper as ERCNN-ONCF framework. The research findings of the paper include a proposed and existing system based on Root Mean Square Error leading to outcomes resulting in a value approximating 0.81 which outperforms other approaches including Positive Matrix Factorization (PMF), Automated Recommender (AutoRec), Restricted Boltzmann Machine (RBM), and Neural Autoregressive Distribution Estimation (NADE). The study also summarizes the performance of the proposed system by comparing the results with the neighborhood approach and probability-based schemes leading to results that were promising.