Scribd
Upload
40 views3 pages

Customer Classification Based On The Historical Purchase Data

The document discusses the importance of customer classification based on historical purchase data to enhance targeted marketing and improve customer retention. It outlines various classification methods, including RFM analysis and deep learning techniques, to categorize customers into groups such as normal, discounted, most profitable, and not a customer. The methodology includes data collection, analysis, model training, and evaluation using specific software and hardware requirements.

Uploaded by

Deepansu Singh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
40 views3 pages

Customer Classification Based On The Historical Purchase Data

The document discusses the importance of customer classification based on historical purchase data to enhance targeted marketing and improve customer retention. It outlines various classification methods, including RFM analysis and deep learning techniques, to categorize customers into groups such as normal, discounted, most profitable, and not a customer. The methodology includes data collection, analysis, model training, and evaluation using specific software and hardware requirements.

Uploaded by

Deepansu Singh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Customer Classification Based on The Historical Purchase Data

Problem Statement
Classification of the customer can be helpful in targeting specific customer for the specific
product. This can remove those irrelevant ads that do not suit for a customer or for which
particular customer will not be interested at all. Also, irrelevant ads floated to the customer might
lead to losing a customer as a customer might irritate of this. Identification of most valuable
person will lead to business luxury. Customer classification is useful for insurance companies for
selling insurance. In financial marketing, it is useful for classifying customer based on the type
of service they used. In banking, it can be helpful in identifying customers and deciding whether
credit card should be issued to a specific customer or whether the loan should be given or to
detect a suspected customer. Customer classification can also be used for dividing network
bandwidth based on customer performance. Such classification will aid in identifying most
profitable customers.
Background
Customers are categorized into different classes based on their purchase history. In food market
customer can be classified based on the type of food they eat: Vegetarian, eggetarian or Non-
Vegetarian, type of food they buy: Uses Fresh vegetable or Frozen vegetable, based on eating
style: Chinese, south Indian, north Indian. In cloth market, they can be classified based on fabric
they used, type of dress they use like: Indian dress, western dress, Indo-westerns. In general, they
can be classified based on time like: regular, sessional, occasional, monthly, yearly or seldom
buyer, based on willingness to buy: tightwads, spendthrift or average spenders. Customers can be
classified based on the risk or product groups. Dividing them based on risk will help in
developing a customer-oriented business approach.
The RFM classification of customers is based on recency, frequency, and money. Recency
means assigning different scores to the customer based on their last purchase period. Assign
more score if the purchase is more recent. Frequency refers to how often customer came and
made a purchase. And money refers to the price of an item a customer buys.
Many visitors visit websites or come at the shop but only some of them made a purchase.
Identifying these parameters will help in deciding the class to which customer belongs.
Identifying the class of customer will provide the detail of future demand. It will also help in
attracting and recommending customer offers that will influence that class customer to buy. Past
studies show that getting a dissatisfied customer back incurs most cost than getting a new
customer and getting a new customer is costlier than keeping an existing customer.
In past smartphone users has been classified based on the application usage sequences. Some has
segmented the customer based on customer lifetime value. Some has used classifier model using
SPSS tool for customer basket data. Some has used Fragmentation-Coagulation Process that
splits during fragmentation and merges the customer groups during coagulation step depending
on their behavior.
The data available for classification is the personal detail of customer, their transaction history,
the frequency of their visit, their click history, their session detail. The commonly used
approaches are HMM (Hidden Markov Model), Decision Tree, EM, SVM based customer
classification.
Methodology
Identifying right customer and providing right service at right time and treating different types of
customers differently is the key to success in business. So, a classifier model will be used to
classify customers into different categories based on their purchase history. Then the model can
be used to predict category of future coming customer. Here, we will use Deep learning
techniques with Keras for classification of the customer.
Step 1 Collecting Customer Data (Purchase History):
This step involves the collection of purchase data from shopping vendors.
Step 2 Analyzing Data and Training Model:
In this step, the collected data will be analyzed and features will be extracted for training the
model. The classifier will be trained for four categories using supervised neural network (RNN)
based method.
Step 3 Testing and Classifying Data (Customer):
In this step, the new tuple (customer) will be classified in one of four categories and will be
tested against the ground truth.

Customer
Classification

Normal Customer Discounted Customer Most Profitable Customer Not a Customer

Evaluation Measures

The performance of a classifier can be evaluated by calculating the accuracy, precision, recall,
specificity.
Software and Hardware Requirements

Software
Anaconda python
Deep learning python libraries
Hardware
The model will be trained using NVIDIA GPUs

Dataset
 https://archive.ics.uci.edu/ml/datasets/wholesale+customers
 https://www.kaggle.com/datafiniti/womens-shoes-prices
 https://www.kaggle.com/kyanyoga/sample-sales-data
 https://archive.ics.uci.edu/ml/datasets/Online+Retail
Reference
[1] Ling Luo, Bin Li et. al., Tracking the Evolution of Customer Purchase Behavior
Segmentation via a Fragmentation-Coagulation Process, Twenty-Sixth International Joint
Conference on Artificial Intelligence (IJCAI-17)
[2] XIANG-BIN YAN, YI-JUN LI, Customer Segmentation based on Neural Network with
Clustering Technique, 5th WSEAS Int. Conf. on Artificial Intelligence, Knowledge
Engineering and Data Bases, Madrid, Spain, February 15-17, 2006 (pp265-268)

You might also like