How to understand user behaviour using Multiple Linear Regression?
summary
Pipeline that turns raw event logs into a single actionable metric using Python code (label encoder, one‑hot/dummy encoding, train/test split) and multiple linear regression with forward/backward (bi‑directional) elimination to pick the highest‑impact feature (p‑value cutoff 0.05). Adds session‑anonymous IDs for GDPR, collects team effort/confidence and smoke‑test clicks to rank feature ROI, and implements a production random forest of per‑user booking trees to serve a personalized “best flight” in real time; works with an ~18‑month travel data decay window.
AI-written summary · about
chapters (12)
- 0:00Introduction to user behavior
- 2:33Understanding our audience
- 7:52Data to knowledge pipeline
- 9:35Multiple linear regression basics
- 13:54Feature importance with p-values
- 16:43Forward and backward elimination
- 18:13Demo of backward elimination
- 21:26Prioritizing features with team data
- 23:18GDPR compliance in data tracking
- 26:16Random forest for flight recommendations
- 30:19How random forest works
- 33:33Q&A session
related talks
29m5 Simple Steps to Create Meaningful Features from Clickstream Data
25mML model from an idea to production with the help...
33mMachine Learning explained to humans
35mCustomer Segmentation in Python
49mUnderstanding Customer Journeys with Deep Learning
15mMachine Learning with Python Case Study
40mMachine Learning over Small Data: generating new variables with predictive power
25mHow we bring Performance Marketing to linear TV by using machine learning