It is a rule-based machine learning technique used to find patterns (relationships, structures) in data.
In Wallmart’s analysis that there is a relationship between diaper and beer sales.
There may be two action ideas here.
- Putting the two products in different places. (sales increase target with products on intermediate shelves)
- Putting products side by side according to lift statistics. (increase bilateral sales)
Association rules in a nutshell; reveal simple statistical patterns, frequencies within the data set.
Achieving these frequencies can give ideas for various actions on store shelves.
It also allows us to use them as recommending systems.
What does it do ?
Contributes to marketing studies.
Contributes to the arrangement of the shelves of the markets.
Supports the design of e-commerce categories.
Cross supports sales. (Suggest a keyboard to the mouse area)
Helps with product discount.
Helps the recommendation system.
The most commonly used Apriori algorithm is used in this repo.
Support: The probability of X and Y occurring together
Confidence: The probability that product Y will be sold when product X is purchased.
Lift: When the product X is purchased, the probability of purchasing Y product increases as much as the lift floor.
Lift is actually a measure of independence.
Here is a measure that gives the independence of X and Y.
**If Lift> 1, it is commented that they are connected to each other and affect each other. **
For example: When milk is taken, the probability of getting cheese increases by lift times.
Sample final table:
Used Online Retail 2 dataset in this project.
This dataset contains the purchase values of a wholesale company's customers in UK between 2010-2011.
InvoiceNo: Unique invoice number. C means refundees.
StockCode: Unique item code
Description: Item description
Quantity: Item quantity number
InvoiceDate: Invoice date time
UnitPrice: Item price (Sterlin)
CustomerID: Unique Customer Number
Country: Country name. The country where the customer lives.
datetime
pandas
pymysql
sqlalchemy
sklearn
lifetimes
Oğuz Han Erdoğan - oguzerdo
VBO - Data Science and Machine Learning Bootcamp
www.veribilimiokulu.com