We implemented three different similarity algorithms basic cosine, adjusted cosine and correlation as described in Section 3.1 and tested them on our data sets. For each similarity algorithms, we implemented the algorithm to compute the neighborhood and used weighted sum algorithm to generate the prediction. We ran these experiments on our train data and used test set to compute Mean Absolute Error (MAE). Figure 4 shows the experimental results. It can be observed from the results that offsetting the user-average for cosine similarity computation has clear advantage, as the MAE is significantly lower in this case. Hence, we select the adjusted cosine similarity for the rest of our experiments.
 |