Anudit Verma
Avid FOSS enthusiast and contributor, interested in UI/UX frameworks, machine learning, AI. Likes listening to music and enthused for gaming. Loves coding and always learning.
Recommendo
A Recommendation Engine cum Data Insight Tool
We Aim To Build
- A tool which analyses customers' behavioural patterns example spending, transaction history, credit score etc.
- It will recommend optimal plans based on the pattern.
-
And provides the bank an understanding about:
- How a package is performing ?
- How it can be improved ?
Methodology
- Create a correlation between user patterns and bank’s policies.
- Classify new user behaviour into existing behavioural patterns.
- When a new cluster of usage pattern comes up as new data is generated, let the bank understand this new demographic.
Attributes that would feed the recommendation engines (prototype):
- Number of products held by customer
- Number of pre-authorized automatic payments
- Credit score
- Customer opening date (to derive years as customer)
- Location information (country, region, state, district, county, zip code, etc.)
- Customer branch associated (where the relationship is held/managed)
- Customer income per annum (it can be an amount in INR or an income band such as INR (1,00,000 - INR 10,00,000)
- Customer total asset value
- Customer group (e.g. bank employee, student, newcomer)
- Customer type (Individual or Organisation/SME Small and Medium Enterprise)
- Customer segment (e.g. Preferred; Platinum/Gold/Silver)
-
Transaction history:
- Merchant category
- Location
Components/Design of the prototype:
- Engine that stores and processes user data.
- REST API for serving results of the engine.
- User profile page (populated with suggestions, plans etc.)
- Admin panel
USE CASE: TAG BSES
- total customer - 10000
- users 1000
- target 90000
- person a
- spends 10000 a month
- given offer
- spends 20000 a month
- plan efficiency 100%
Product Design
Tech Stack
- Python (NumPy For GaussianNB classification)
- MongoDB
- Flask (RESTful API)
Zafin Tech
By Anudit Verma
