Build a module to classify 5-star or others. Tried various combination of hyper parameter and neural network parameters.
We have selected 500 features and used TF*IDF weight and seed value as 12345 1. Two class neural network with default parameter gives the accuracy of 58% 2. By tuning the hyper parameter, we have selected random grid and accuracy improved to 64% 3. We tried with 2 hidden layers 1. Here both the layers have default number of nodes and our accuracy came out to be 63.9% 2. Here I have selected 75 nodes in first layer and 150 nodes in second layer. Changed learning rate to 0.15 and learning iteration to 200 and our accuracy came out to be 64.1% 4.We have build the recommendation system using the variable reviewID, asnid(Product_ID), overall and the model recommends the item to the user. We didn’t give any input to user feature and item feature. User feature is used to give attributes regarding user and Item feature is used to give attributes regarding item. NDGC came out as 97.5. 5.We executed R script to build the word cloud with 100 words.