Mrudula Jujala

Estimate E-commerce Customers buying behavior with Machine Learning /AI

Introduction:

E-commerce is the buying and selling of good or services via the internet, and the transfer of money and data to complete the sales. It’s also known as electronic commerce or internet commerce.
E-commerce service is for every one. Electronic business draws on technologies, for example, mobile commerce, electronic funds transfer, supply chain management, Internet marketing, online transaction processing, electronic data interchange (EDI), inventory management systems, and automated data collection systems. Electronic commerce is mostly regarded to be the sales prospect of e-business.. It also comprises of the trade of information to promote the financing and installment parts of business transactions. This is an effectual and efficient means of communicating within an organization and one of the most efficient and utilitarian ways of carrying on job.

Aim: The main aim of the study is to analyze Influencing & Motivating Factors for Customer’s buying behavior. Identifying the influence of demographic variables on customer’s buying behavior. To implement the work, multiple efficiency parameters such as psychological, technological, informative will be added.

PHASE I

Objective:

The objective is to predict if the person is reliable or not for an E-Commerce website. This can be said by performing analysis to various attributes like Age, Educational Qualification, Gender, Employment status, etc.
So this analyzed data will then be compared to featured data which is obtained from surveying(where we predict several demographics data like age, gender ), then it is compared with the analyzed data and the client gets to know if the customer is reliable to his website or not.

About the data set:

There is data of 2,00,000 existing Customers information which is available on SQL Metadata Server, with corresponding details.
We blasted through our Web app in various Google SEO, Digital marketing Apps, and social media sites for gathering more information .
Then from the response we received around 2909 samples randomly from the users.
There are 43 variables, among which there are some attributes that contains the demographic data of the users.

Dataset: https://drive.google.com/file/d/1U1b19ehbtNBLPYdSa9HzbW9k1qV4Rctu/view?usp=sharing

Methodology:

Our data was structured data and target variables are demographic and psychographic variables. So we are using classification Machine learning algorithms. Before Model implementation, we are doing data dimension reduction which comes under Unsupervised Machine learning.
The performance will be measured based on confusion matrices, RMSE, APE, AIC BIC.
Can solve the problem manually through comparison of models we can do this ex AIC BIC.

Results from the testing data will be analyzed using Data mining Tools and Techniques which are expected to be more effective for analyzing consumer buying behaviors.
To present groups of clients with different features and to position the target clients for e-commerce, Need to select appropriate techniques for mining, summarizing, visualizing data.
The following are the statistical phases to analyze the data
Regression
Classification.

PHASE II

The research questions and hypothesis were developed on a review of the literature. The research was carried out on selected 500 customers using E-Commerce applications in India. In this research, the survey method was selected to understand the factor affecting customer attitudes towards E-commerce applications in India. For data collection, random sampling was

adopted. To ensure all questions being answered in a proper way, questionnaires were completed and screened one by one. Data mining Tools and Techniques i.e. Weka, R programming, R Rattle, and SPSS 20 software (Regression, Classifying& Clustering) has been used to analyze the data. This study helps to analyze the pattern of online buying (types of goods, e-commerce experience and hours use on the internet) influence customer attitude towards online shopping.

DATA PREPARATION:

To measure these objectives a cross-sectional descriptive study was designed.

Questionnaire Design: Based on the research objectives, a structured questionnaire with 30 variables, mainly with a 5-point Likert scale was used, in which 1 = strongly disagree and 5 = strongly agree.

Data Collection: In this research, the survey method was selected to understand the factor affecting customer attitudes towards E-commerce applications in India. For data collection, random sampling was adopted. To ensure all questions being answered in a proper way, questionnaires were completed and screened one by one.

Data Accessing and Cleaning: The data collected through the survey is saved in CSV format. This data is converted into data table format. The data cleaning and validation process are carried out using the R Programming language and Weka tool.

Data wragling Cleaning: After collecting the data, Data quality validation has been done Python SPSS 20, R programming.
Analysis Tools & Techniques: Data mining Tools and Techniques i.e. WEKA, R programming, R Rattle, and R Rcmdr have been used to analyze the data. Multinomial Logistics model is applied for predicting customer buying behaviors. Data mining techniques like Principal Component Analysis and factor analysis are applied for Data Dimension Reduction. Originally there were 30 factors influencing attributes/variables which were reduced to 7 after the PCA technique is applied. And the analysis is carried here afterward with the reduced 7 variables.

Data Validation and Quality Check:

Data is validated by performing the reliability test for the data, using Cronbach’s Alpha. KMO and Bartlett’s tests are performed for sample adequacy. A good result of 87% is observed indicating the nmnreliability percentage of the data. Also, the KMO test for the factor data validation is obtained to be 96%. Tables for the data reliability and validity tests are given below.

Validation Imputation of PCA Factors:

PHASE III

MODEL ACCURACY:

Marital Status : Model Validation summary report [Binary Model] 67 %

Customer Gender : Model Validation Summary report [Binary Model ] 63.3%

Customer Region : Model Validation summary report [Multinomial Regression] 90.1%

Income Group :Model Validation summary report [Ordinal regression ] 83.63%

Customer Occupation : Model Validation summary report [Logit Regression ] 66.85%

Education: Model Validation summary report [Ordinal Regression ] 84.59%

Age Group: Model Validation summary report [Ordinal Regression ] 60.52%

Limitation:

A limitation of this study lies in the sample being from one place. An additional, more diverse, the sample should be examined. Also, this study did not seek to address influences on trust and risk. Many additional factors could be proposed as influences on trust and risk and should be studied in the future. In particular, one’s experience level or a number of years of involvement in e-commerce could be used to explain one’s attitude toward buying and/or selling. This study also measures perceptions and not true behaviors. Therefore, future studies should seek to gather data regarding actual consumer behaviors in e-commerce. Finally, given the combination of trust and attitude in the attitude toward selling model, further research is needed to examine why a consumer’s attitude and trust might be combined in certain situations or environments.

Conclusion:

The results of this study will provide insightful information on customer attitudes towards E-Commerce in India. The result identifies major problems of Customer behavior and their direct impact on the sale and production. Although this study is based on the Indian scenario of business, besides this the results derived from studies carried out in the field are equally applicable in other developing countries. This study is mainly focused on how demographic variables (age, income, and occupation) affect customer attitude towards E-Commerce. This study helps to analyze the pattern of online buying (types of goods, e-commerce experience and hours use on the internet) influence customer attitude towards online shopping. This research focuses on how purchase perception (product perception, customer service, and consumer risk) influences customer attitude towards E-Commerce. This research has identified that there are a number of factors determining a customer’s intent to repurchase within e-commerce, specifically, these factors are customer experience with an e-brand and beliefs concerning the importance of convenience, trust, and security when purchasing.

GitHub: https://github.com/mrudulajujala/Customer-buying-behavior-

References :

1.Hou Zubing, Yu Guohe “An Empirical Research on Influence Factors of Online Shopping”, 1st International Conference on Information Science and Engineering (ICISE),pp.2846-2849, 2009

2.Kang Boyu “The Empirical Study of Motivators for Online Shopping Demand Based onBehavior Analysis”, Second International Conference on Business Computing and Global Informatization (BCGIN), pp. 225-228, 2012

3.Bakar, E.A., Lim Li Chang ; Saidin, A.Z. “Knowledge, attitude and practices of consumers in e-commerce transactions”, 5th International Conference on Information and Communication Technology for the Muslim World (ICT4M), pp.1-7, 2013.

4.Hengjun Zhao, Changzheng He “Objective Cluster Analysis in Value-Based Customer Segmentation Method”, Knowledge Discovery and Data Mining WKDD, pp.484-487, 2009.

5.Mahdavi Anari, S.M., Bakri, A. ; Ibrahim, R. “Understanding factors on the customer intention behavior through Facebook commerce: A conceptual model”, International Symposium on Technology Management and Emerging Technologies (ISTMET), pp. 107-112, 2014.

6.Nanang Husin, “Internet User Behaviour Analysis in Online Shopping on Indonesia”, ICACSIS, pp. 137-142, 2011.

Page updated

Report abuse