Analysis-and-prediction-of-online-shoppers-purchasing-intention-using-various-algorithms-CAPSTONE

In this project, we aim to construct a real-time prediction machine learning system for online shopping environment. We use an online retailer data to perform the experiments. In order to predict the purchasing intention of the visitor, we use aggregated page view data kept track during the visit along with some session and user information as input to machine learning algorithms. Oversampling/Undersampling and feature selection pre-processing techniques are applied to improve the success rates and scalability of the models.

Our findings support the argument that the features extracted from clickstream data during the visit convey important information for online purchasing intention prediction. The findings show that choosing a minimal subset of combination of clickstream data aggregated statistics and session information results in a more accurate and scalable system.

Keeping last mile goal of having predictive model in mind, we used different data cleansing, data balancing imputation methods to arrive at dataset fit for development of predictive models. The project includes creation of various analytical models for predicting and classifying users as Revenue or Non-revenue generating using Logistic Regression, Ensemble methods, Bagging, Boosting techniques etc. The base models are benchmarked against Hyperparameter tuned model outputs and considering all factors and final goal in mind best model is advised.

Techniques: Predictive Modelling, Machine learning, Hyperparameter tuning Tools: Python, Tableau Domain: Marketing and Retail Analytics

Analysis shows that it is possible to predict a site visitors buying behavior within a certain level of confidence based on the features we were able to pull from the data set. Considering the real time usage of the proposed system, achieving better or similar classification performance with minimal subset of features is an important factor for the e-commerce companies since a smaller number of features will be kept track during the session. XGBoost model turned out to be the best model overall.

Name		Name	Last commit message	Last commit date
Latest commit History 9 Commits
1-Imputation_Techniques_ContinuousVariables.html		1-Imputation_Techniques_ContinuousVariables.html
2-Imputation_Technique_Categorical+Variable.html		2-Imputation_Technique_Categorical+Variable.html
3-Engineering+Outliers.html		3-Engineering+Outliers.html
5-Dealing+with+Categorical+Variables.html		5-Dealing+with+Categorical+Variables.html
7-Engineering_Dates.html		7-Engineering_Dates.html
Capstone Presentation_CPGP1_FINAL.pdf		Capstone Presentation_CPGP1_FINAL.pdf
CapstoneFinalReport_CPG1APR19_v1.3.pdf		CapstoneFinalReport_CPG1APR19_v1.3.pdf
Cat_Var Handling_CorrMatrix_and Simple models.ipynb		Cat_Var Handling_CorrMatrix_and Simple models.ipynb
ConstantFeatures.ipynb		ConstantFeatures.ipynb
NAs_Outliers_NegativeValues_Imputation.ipynb		NAs_Outliers_NegativeValues_Imputation.ipynb
Prelim 15_Classification models with SMOTE-0.6oversampling.ipynb		Prelim 15_Classification models with SMOTE-0.6oversampling.ipynb
Prelim 15_Classification models with SMOTE.ipynb		Prelim 15_Classification models with SMOTE.ipynb
Prelim 15_Classification models with SMOTeenOverUndersampling.ipynb		Prelim 15_Classification models with SMOTeenOverUndersampling.ipynb
Prelim10_Encoding+Scaling_Ft.select_HybridRFE.ipynb		Prelim10_Encoding+Scaling_Ft.select_HybridRFE.ipynb
Prelim11_Encoding+Scaling_Ft.select_HybridRecursive feature addition.ipynb		Prelim11_Encoding+Scaling_Ft.select_HybridRecursive feature addition.ipynb
Prelim12_Encoding+Scaling_Ft.select_GradBoostImportance.ipynb		Prelim12_Encoding+Scaling_Ft.select_GradBoostImportance.ipynb
Prelim13_PCA_Check.ipynb		Prelim13_PCA_Check.ipynb
Prelim14_NB+DT+SVM+KNN_LogReg+Ensemble.ipynb		Prelim14_NB+DT+SVM+KNN_LogReg+Ensemble.ipynb
Prelim1_ProperEncoding_AllBasemodels(lg,rf,dt,NB,KNN,LinearSVC,AdaBoost,GradientBoosting,xgboost).ipynb		Prelim1_ProperEncoding_AllBasemodels(lg,rf,dt,NB,KNN,LinearSVC,AdaBoost,GradientBoosting,xgboost).ipynb
Prelim2_ProperEncoding+ScalingNumVars_AllBasemodels(lg,rf,dt,NB,KNN,LinearSVC,AdaBoost,GradientBoosting,xgboost).ipynb		Prelim2_ProperEncoding+ScalingNumVars_AllBasemodels(lg,rf,dt,NB,KNN,LinearSVC,AdaBoost,GradientBoosting,xgboost).ipynb
Prelim3_ProperEncoding+ScalingNumVars_Correlationcheck&drop.ipynb		Prelim3_ProperEncoding+ScalingNumVars_Correlationcheck&drop.ipynb
Prelim4_Encoding+Scaling_Ft.select_InfoGain.ipynb		Prelim4_Encoding+Scaling_Ft.select_InfoGain.ipynb
Prelim5_Encoding_Ft.select(CatVars)_FisherScore.ipynb		Prelim5_Encoding_Ft.select(CatVars)_FisherScore.ipynb
Prelim6_Encoding+Scaling_Ft.select_Univariate+ROC_AUC.ipynb		Prelim6_Encoding+Scaling_Ft.select_Univariate+ROC_AUC.ipynb
Prelim7_Encoding+Scaling_Ft.select_Stepforward&backward&EFS.ipynb		Prelim7_Encoding+Scaling_Ft.select_Stepforward&backward&EFS.ipynb
Prelim8_Encoding+Scaling_Ft.select_rfFeatureImp_rfRFE.ipynb		Prelim8_Encoding+Scaling_Ft.select_rfFeatureImp_rfRFE.ipynb
Prelim9_Encoding+Scaling_Ft.select_FeatureShuffling.ipynb		Prelim9_Encoding+Scaling_Ft.select_FeatureShuffling.ipynb
README.md		README.md
Visual EDA.ipynb		Visual EDA.ipynb
featureselectionsheet_modelresults.xlsx		featureselectionsheet_modelresults.xlsx
online_shoppers_intention_1.csv		online_shoppers_intention_1.csv

Aviator10/Analysis-and-prediction-of-online-shoppers-purchasing-intention-using-various-algorithms-CAPSTONE

Folders and files

Latest commit

History

Repository files navigation

Analysis-and-prediction-of-online-shoppers-purchasing-intention-using-various-algorithms-CAPSTONE

About

Topics

Resources

Stars

Watchers

Forks

Languages