An Algorithmic Approach for Mining Customer Behavior Prediction in Market Basket Analysis

Abstract

Market basket analysis is the search for data that contain customer purchasing items. Market basket analysis is a process of showing the correlation between the data with respect to support and confidence. Support indicates that how frequently items appear in the database and confidence indicate that rules must be generated based on the frequent items. Data analysis in a supermarket database means to understand each transaction available in the dataset that contains customer purchasing pattern to determine how the product should be located on shelves. Product arrangement is the most important aspect to get supermarket profit. The dataset of the retailer contains transaction of the items which is purchased by the customer and also comment regarding that product whatever they fill regarding that product. Apriori-algorithm used to find frequent items and association rule based on customer transactions. Frequent items calculate with respect to support and Association rule determines with respect to confidence. This paper tells about how customer behavior predicted based on the customer purchase items. This technique generally used in the agricultural field, marketing, and education field.

Keywords: Data mining, Market Basket Analysis, Customer behavior, Apriori Algorithm, Association Rule, Layout, Support, Confidence

Introduction

Market basket analysis is one of the techniques that analyze customer purchasing habits by finding the different relationship between the different items that can be stored in customer shopping baskets. Association rule can help retailers to produce effective marketing strategies by gaining items frequently, purchased together by customers. Data mining is the understanding of large datasets to find the irrelevant association and summarize the data in proper ways both are understandable and useful to the retailer. Knowledge discovery database is discovering informative knowledge from a large amount of complex data. The knowledge discovery database is a process of interactive and iterative data form from the large database. It contains different steps such as selection, preprocessing, transformation, data mining and interpretation or evaluation. Each step performs their own role to discover informative knowledge from the database.

Market basket analysis is an example of elaborating association rule mining. It is one of the technique that all the retailer in any kind of shop or departmental stores would like to gain knowledge about the purchasing behavior of every customer. These results help to guide retailer to make a plan for marketing or advertising approach. market basket analysis will also help managers to propose a new way of arrangement in store. Based on this analysis, items are regularly purchased together that can be placed in close proximity with the purpose of further promoting the sale of such items together. If consumers who purchase computers also likely to purchase anti-virus software at the same time then placing the hardware display close to the software display will help to enhance the sales of both of these items.

Market basket analysis is an example of extracting association rule mining. It is a fact that all the managers in any kind of shop or departmental stores would like to gain knowledge about the buying behaviour of every customer. Association rules are if-then statements that help to uncover the relationship between seemingly unrelated data in a relational database or other information.

Related Work

The work in describe the support and confidence have been calculated by the generic formulae and it does not give the complete information of the association rule. A Database containing all the transaction of items.

The researchers in describe the product which is the relationship with one another finding with the help of market basket analysis is located in the store layout.

In another survey authors, Information system containing the relation between each customer purchasing item that is helpful to get the future decision.

The work in describes it used in sports company regarding purchasing the sports items through the customer. It identifies purchasing pattern of sports items which is present in the database.

Researchers discovered, Market basket analysis is used to discover purchasing patterns of customers by extracting associations from different store transactional data.

Proposed Approach

Dataset

The dataset is a relational set of files describing customer's orders. The input data for a Market Basket Analysis is normally a list of sales transactions where each has two dimensions one represents a product and the other represents a customer.

Data Preprocessing

All the items in the transaction are sorted in descending order in reference to their frequencies. The algorithm does not depend upon the specific order of the frequencies of items sorting in descending order may lead to much less execution time than ordered randomly.

Apriori Algorithm

Apriori algorithm generates sets of large items-sets that find each support size of items. The complexity of an apriori algorithm is always high. Frequent item-sets are extended one item at a time and group of candidates is tested against data. It operates all the transaction which is present in the database.

Findings

Input: Database containing items Output: Frequent Item-sets

Algorithm

1. S is a dataset containing the item. Minimum support is less than 1 and greater than 0. Minimum support is real.
2. Take a transaction of the customer.
3. Calculate support for each item.
4. Take the first transaction and so on.
5. Calculate support for the first item which is the ratio of the number of transaction containing the item and a total number of the transaction.
6. Compare item support with minimum support. Item support is greater than or equal to the minimum support.
7. It generates frequent item-set
8. Again go to Step 4 and calculate all frequent item set.

Association Rule

It contains if-then rules which support the data. Market basket analysis is an association rule which deals with the content of point-of-sale transaction of large retailers. It identifies the relationship among the attribute which is present in the database. It assigns relationship of one item with another item. It is a fact that all the managers in any kind of shop or departmental stores would like to gain knowledge about the buying behavior of every customer.

Frequent Item-set

There are ‘n’ items and it gives multiple combinations of ‘n’ items and at last, the customer selects the proper combination of items according to their own choice.

Customer Behavior

Market Basket Analysis allows retailers to identify relationships between the products that people buy. Targeting market must send promotional coupons to customers for products related to items they recently purchased. Most of the customer buy the same product according to their requirement.

Classification Rule Mining

Classification rule mining is an effective strategy to generate customer behaviour. Obtaining comprehensible classifiers may be as important as achieving high accuracy that help to make an effective decision in business. The complexity of classification rule mining is O(pCp n ) where, p is number of items in classification rule and n is number customer transactions. It contain more than three items to generate the rules.

Classification method is better than the association rules. Dataset contain different fields of items purchasing by the customer. Association rule contain two items so it cannot give complete count based on whole dataset instead of that classification rule mining give count based on whole dataset using different field of items.

Result

Association rule mining generated from the apriori algorithm. The rule must be generated according to that rule we get customer behavior. Customer behavior based on the rule generated from the dataset. Dataset containing items purchased by the customer with their quantity, unit price and so on. This rule contain 2 items. Based on the rules graph must be generated that shows smaller and larger circle contain large or smaller amount of customer purchasing such items. Together how many customer purchasing those item based on rule must be generated hence it gives limited count as compare to classification rule.

Classification based association which is generated by the association rule. It gives better performance than the association rule. It reduces the complexity of apriori algorithm and improves the performance. Classification rule mining is better as compare to association rule mining.

The customer behavior generated from the classification rule. This rule generated from the country so it gives actual count based on complete dataset.

Algorithms Accuracy

Association Rule Mining 74.58 %

Classification Rule Mining 81.33 %

Conclusion

This paper shows that market basket analysis is an important tool to get frequent item and relationship between the items. It generates the frequencies of the item. Based on their frequency item placed into the store layout. The item that is frequently purchased by the customer that is first placed on the layout. Items placed one after another that are helpful to the customer to search the items easily. Apriori algorithm help to generate association rule and frequent item-set. The apriori algorithm help to get association rule mining algorithms for market basket analysis will help in better classification of the huge amount of data. The apriori algorithm can be modified effectively with respect to classification rule mining that reduces the time complexity and enhances the accuracy. Association rule help to get customer behavior based on the product. Both association rule and classification rule able to get the customer behavior based on products different field.