Recommendation-Engine-using-R

Recommendation Engine:

It works based on Collaborative Filtering

How Collaborative Filtering Works:

  step 1: Preprocessing before building matrix
  step 2: Build a matrix between customers and products
  step 3: Find similarity between customers
            we have 2 methods:
              1. cosine based similarity
                          cos(A,B) = A.B / (|A|.|B|)
              
              2. correlation based similarity
                            corr(A,B) = Covariance(A,B)/(std(A)*std(B))
   step 4: once we find similarity items will be recommended

Negatives of Collaborative Filtering:

1. Memory based - as it needs a huge amount of matrix with data
2. As big matrix computationally heavy

How to reduce Computation

1. Randomly sample customers
2. Discard infrequent buyers
3. Discard items that are very popular or unpopular
4. clustering can decrease number of rows
5. PCA can reduce number of columns

Data used:

Movies recommendation is done using Movie dataset

Programming:

R Programming

The Code regarding Movie recommendation and its dataset is present in this Repository in detail