Recommendation Engine

As name says, it is process data for Hybrid Recommendation. We have used multiple things like rating, review, popularity and similarity for better recommendation.

Minimum required data

product_info table with :
'prodId','name','description'

rating_review_info table with:
'userId','prodId','product review','rating'

user_history_info table with:
'userId','prodId'

Project Set Up and Installation

1. Clone the repository from this URL: [http://192.168.30.240/poojan.shah/recommended-engine]
2. Create Virtual Environment in working directory.
   cd recommended-engine
python3 -m venv venv
3. Activate Virtual Environment
   source venv/bin/activate
4. Set re.config file for Data Connector
   dtype: main input data source type (fake, csv, mysql, sqlite)
   if dtype : csv
   prod_info_path: path of prod_info file
   rat_info_path: path of rat_info file
   uhis_info_path: path of uhis_info file
   if dtype = mysql
   user: user
   password: password
   host: host
   if dtype = mysql or sqlite
   database: name of your database
   table names
   prod_info: tablename (who have minimum product id, name and description columns)
   rat_info: tablename (who have minimum user id, product id, rating and review columns)
   uhis_info: tablename (who have minimum user id and product id columns)
   column names
   prodId: column name of products' id of your data
   name: column name of products' name of your data
   description: column name of products' description of your data
   userId: column name of users' id of your data
   rating: column name of product's rating of your data
   product_review: column name of product's review of your data
   others
   only_qualified: False (only_qualified product will use based on the minimum rating count required to be listed in the data.)
   sdb: database name (for saving processed data and use for API)
   Caution: every variable in each line with appropriate manner (example: variable:<single_space>variable_name )
5. Just execute main.py (for recursive process set worker for this)
   python3 main.py
6. All is ok, then you are good to Go for APIs!!!

For APIs

visit this URL: documentation for APIs

Logic behind the beast

Fake data creation
Created fake data using Faker library

Popular products
find visitor count per product and normalise

Weighted Average
rating average and find score based on given formula
Formula: (v/(v+m) * R) + (m/(m+v) * C)

---

v is the number of ratings
m is the minimum rating count required to be listed in the data
R is the average rating of the product
C is the mean vote across the whole data.

Sentiment of text review
used pre-train model of twitter of hugging face
sentiment and group by fresult per product
https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment/tree/main

Content based filtering
Word2vec vectorization and nearest neighbour used for finding similer product

Sorting best product for user
similar product sort based on weights 1] sort based on user visited count * neighbour distance(20%)
2] popular product (10%)
3] sort based on rating (30%)
4] sort based on text positive review (40%)

Flow diagrams

flowchart LR
A[(Shopping DB)]<-->B[Web Portal]
B[Web Portal]--Send User's History Product Ids-->C[Recommendation system]
C[Recommendation system]-.Share Recommended Product Ids with score.->B[Web Portal]
C[Recommendation system]<-- Get Relevant information based on IDS -->D[(Local DB)]
A[(Shopping DB)]--Get needed data fields-->E[Sync DB Process]
E[Sync DB Process]-- Post Process the available data-->D[(Local DB)]

Loading