A Network-based Recommendation Algorithm via Improved Similarity Model

摘要

Many researchers have devoted their works toward improving the effect of recommendation algorithms. Here a new method is introduced to recommend information to users based on the Improved Similarity Model (ISM). Through the use of the well-known data set MovieLens as test data, the experiments testify that this method has a good effect on recommendation than other methods. The method can achieve the optimal value when the parameter in the ISM formula equals to a special value. By comparing the ISM model with several traditional models, the results show that the ISM model always has best recommendation effect in different test criteria fields. This model can significantly outperforms traditional models by not only enhancing recommendation accuracy but also improving recommendation diversity and giving more personalized recommendations.