Walmart Stores Sale Forecasting

Project Overview

Our project focuses on sales forecasting and how certain outside influences can affect store performance. We will predict sales for Walmart based on three years of data across 45 stores by using various machine learning models. We want to determine which model generates the most accurate predictions.

Project Team and Role Responsibilities

Segment 1

Archana Rohilla - Created preliminary Machine Learning Models, Imported data from RDS Database "Walmart sales" into Jupyter Notebook

Vick Anand - ETL analysis, build schema utilizing ERD QuickDB, Postgress and connection to AWS RDS DB

Tri Luu - Presentation and communication plan (readme.md)

Frances Klingenberger - created github master branch and tracked communications

Segment 2

Archana Rohilla - Imported data from RDS Database "Walmart sales" into Jupyter notebook, Expanded Machine Learning Models using "lags"

Vick Anand - Created Dashboard with Tri

Tri Luu - Created Dashboard with Vick

Frances Klingenberger - Manage Github and expanded Presentation

Segment 3

Archana Rohilla - Expanded Machine Learning Models using "Naive Forecasting Technique" with Vick

Vick Anand - Created Dashboard with Tri and worked on Machine Learning Model with Archana

Tri Luu - Created Dashboard with Vick

Frances Klingenberger - Manage Github and expanded Presentation

Segment 4

Archana Rohilla - Finalized Machine Learning Model with Vick, Practiced Presentation

Vick Anand - Finalized Dashboard and Machine Learning Model with Archana, Practiced Presentation

Tri Luu - Finalized Dashboard and Presentation, Practiced Presentation

Frances Klingenberger - Manage Github and Finalized Presenatation, Practiced Presentation

Process Overview and Technology used

ETL

Extract: Extracted csv file from Kaggle.com

Transform: We created an ETL function to revise and clean columns

Load: Loaded data into Postgress from Jupyter Notebook

Source: ETL/Walmart_Wkly_Sales_ETL.ipynb

Machine Learning Model:

See full machine learning description in the Machine_learning_Report.pdf

Model Type: Supervised Machine Learning : Linear Regression and Random Forest Regression

Why we are using this Model: To predict Walmart stores sales for 45 stores in aggregate utilizing features like temperature, holiday, CPI, fuel price and unemployment data.

How are we training our Model: We used Linear Regression and Random Forest Regressor to fit the training dataset, scale and transform the data to predict sales data. After that, we compared the test data actual with the predicted value.

How does this Model work? We used 3 metrics to compare the effectiveness of Machine Learning Models. The 3 metrics are: RMSE, MAE, and R2 score.

• RMSE: The square root of the average of squared differences between the predicted values and actual values.

• MAE: The average of the absolute differences between the predicted values and actual vales, all differences have the same weight.

• R2 score: A measure of best fit that shows how much variation of a dependent variable is explained by the independent variable.

• Naive RMSE: Same as RMSE with the assumption that predictions will mirror previous week sales.

Database

Develop database schema utilizing ERD DB below:

Sources: ETL/QuickDBD-WMT_Weekly_Sales.sql

Database Connections & Steps: database/Steps for Database & Connections.docx

RDS DB endpoint: walmartsales.ctixdh2hiprk.us-east-2.rds.amazonaws.com

Dashboard

Links to Dashboard:

https://public.tableau.com/profile/vick.anand#!/vizhome/Walmart_Sales_ML_Prediction-revised/WalmartSalesOverview

https://public.tableau.com/profile/vick.anand#!/vizhome/Walmart_Sales_ML_Prediction-revised/MachineLearningResults

Final Presentation

Link to Presentation/Dashboard: https://public.tableau.com/shared/NBPGWZXQT?:display_count=y&:origin=viz_share_link

Link to Presentation Notes: https://docs.google.com/document/d/1PgImt_HTl0jHM45ZBqyobfUrsQerVCWuR3aXJnAJICE/edit?usp=sharing