Data science with Google Analytics @MeasureCamp
5 likes3,164 views
The document discusses the integration of data science with Google Analytics, emphasizing that it is indeed possible, provided one accesses data via Google API and utilizes R/Python libraries. It outlines a step-by-step approach, including data transformation, storage, and modeling techniques such as decision trees and clustering. The author encourages leveraging algorithms and enriching data through internal and external sources to effectively utilize Google Analytics for data science.
Data science with Google Analytics @MeasureCamp
- 1. Data Science with Google Analytics Is it possible and how ? Alexandros Papageorgiou MeasureCamp London. September 2017.
- 2. About me Account strategist @ Google Back to school - Data analytics analyst @alex-papageo.com Consumer behaviour analyst @ WhatClinic.com
- 3. Data...data...data It’s in our menu every day! Is it Data science ? Can it be Data science ? (what’s data science anyway ? )
- 4. “Google Analytics is basically for marketers” Data Scientist in big internet company to me
- 5. The landscape back in 2015
- 6. A year or so later
- 7. Hype ?
- 8. Back to the question: Is DS with GA possible ? Answer: Yes it is. But…
- 9. 1. Access the Data Google API + R/Python Libraries. R GoogleAnalyticsR RGA RGoogleAnalytics Python Pandas ga google2pandas
- 10. 2. Initial variable selection
- 11. 3. Un-sample the data E.g. Day by day or hour by hour api calls.
- 12. 4. Transform the data Bring it to atomic/event level, e.g. every row corresponds to an individual obervation - User - Session - Page
- 13. User explorer ?
- 15. Result: Session level data
- 16. Custom dimensions Supercharging websites with a real-time R API http://code.markedmondson.me/predictClickOpenCPU/supercharge.html#1 Improve Data Collection With Four Custom Dimensions https://www.simoahava.com/analytics/improve-data-collection-with-four-custom-dimensions/ E.g. Cust ID + Timestamp
- 17. 5. Store the data (recommended)
- 18. 6. Model & Communicate the data https://github.com/papageorgiou/dublinR-talk-analytics http://www.alex-papageo.com/research.html E.g. Decision tree & variable importance for conversion prediction
- 19. Some ideas ● Decision tree for conversion prediction (rpart) ● Clickstream analysis to predict next page (clickstream) ● Clustering for customer segmentation (base R) ● Association Rules for pages/products frequently visited together (arules) Enrich data: Look for opportunities to join with internal data/ external api data
- 20. Wrapping up DS with GA not straightforward but possible Take advantage of GA API Open source R/Python libraries Get familiar with 2-3 algos and apply them on your data.
- 21. Thank you!