Mod 2 Project: Food's Gold RFP

Tom Hill

Project:

A general store named "Food's Gold" has approached Ph[0]ton to perform an analysis on their marketing data to produce insights as to which demographic to target for their next campains.

They provided a dataset consisting of marketing data from 2012-2014 for me to explore and generate recommendations.

Objective:

• Create a linear regression model that can predict the amount a customer will spend on wine over the next two years.

Notes:

• Food's Gold is a fake company
• The dataset came from Kaggle and can be found here: https://www.kaggle.com/rodsaldanha/arketing-campaign
• The presentation assumes the year is 2014 and that we are identifying opportunities for Q1 2015
• These assumptions do not and did not change the results of my analysis - instead used as a storytelling device

Process:

• Data cleaning consisted of:

1. adding missing income values with the median
2. engineering features for Education, Generation, and Marital Status
3. producing dummy variables for One-hot encoding

• Statistical tests included:

1. Anova
2. Welch's Ttest

• Linear Models used:

1. Standard
2. Polynomial
3. F-Test
4. Recursive
5. Lasso

Libraries used:

• NumPy
• Pandas
• Matplotlib
• Seaborn
• SciPy
• Statsmodels
• Scikit Learn

Results

*Original Training Error: 172.02446760608765

*Original Testing Error: 196.58156017150185

*Train R2 0.7328327537833788

*Test R2 0.6850644337462928

---

*Poly Training Error: 87.43126820497908

*Poly Testing Error: 3425072180673.2207

*Train R2 0.9309861389565699

*Test R2 -9.560390038500187e+19

---

*F-Test Training Error: 176.68838666766968

*F-Test Testing Error: 203.70550540540967

*Train R2 0.7181495194164935

*Test R2 0.6618248533061816

---

*Recursive Training Error: 171.87208883649276

*Recursive Testing Error: 195.84138522503287

*Train R2 0.7381359055139782

*Test R2 0.6661798156645562

---

*Lasso Training Error: 115.63858719809843

*Lasso Testing Error: 195.7476059005185

*Train R2 0.7381353445435714

*Test R2 0.666222001860492

Winner:

• The Recursive model produced an RSME Z score of 0.52

Technical Documentation can be found in the following notebook:

• Mod 2 Project - Food's Gold Analysis

Presentation can be found here:

• https://github.com/photonconsult/DS_51120_Mod_2/blob/master/Foodsgold__Photon_Presentation.pdf