Dacker15 / r-reccomendation

An overview of reccomendation systems in R

Recommendation Systems in R

In this project, we compared the results of different recommendation systems using the recommenderlab library based on the MovieLens Small dataset.

The recommendation systems used are based on two main types:

• UBCF (User Based Collaborative Filtering)
• IBCF (Item Based Collaborative Filtering)

trained on the same dataset, but separated with different techniques:

• Split
• Bootstrap (sampling with replacement)
• Cross Validation

with different comparison measures between items:

• Cosine Similarity
• Pearson Correlation

and with different measures for normalizing items:

• Center
• Z-Score

The aforementioned recommendation systems were also trained on a dataset where the average ratings are weighted using the timestamps of individual ratings.

Additionally, a binary recommendation system was implemented, where the values are converted from the scale $[0, 5]$ to the scale $[0, 1]$, using a threshold value of $3$. For this system, the only comparison measure is Jaccard similarity and there are no normalization metrics.

The complete report, which includes a comprehensive analysis of the data and a detailed explanation of each recommendation system, is available as a PDF inside the repository. The report is written in Italian.