GEOG5917 – Big Data and Consumer Analytics

GEOG5917 – Big Data and Consumer Analytics

University of Leeds | School of Geography

🎯 Course Overview

This module explores the concepts, techniques, and applications of big data in business analytics. Students will learn about data linkage, model building (including spatial and machine learning models), reproducibility, and ethical considerations in big data analytics. Through practical assignments, students will develop skills in predictive modeling using machine learning approaches, particularly Gradient Boosting Machines (GBM).

📅 Weekly Lecture Themes & Content Focus

  • Week 1: Introduction to Big Data
    • What is Big Data?
    • Introduction to CDRC, LIDA and applied Big Data research
    • Importance of Big Data to business analytics
  • Week 2: Data Fundamentals
    • Data linkage techniques
    • Data summarization methods
  • Week 3: Making Models
    • Model building process
    • Regression models (OLS, GLM)
    • Spatial considerations in modeling
  • Week 4: Machine Learning Fundamentals
    • Machine learning vs classic statistics
    • Prediction vs inference
    • Training and validation data
  • Week 5: Advanced Modeling Techniques
    • Model tuning and evaluation
    • Introduction to Gradient Boosting Machines
  • Week 6: Spatial Models
    • Tobler’s First Law of Geography
    • Geographically Weighted Regression (GWR)
    • Multi-Scale GWR (MS-GWR)
  • Week 7: Reproducibility and Transparency
  • Week 8: Big Data Privacy and Ethics
  • Week 9: Future of Big Data
    • Evaluating the hype around big data
    • Future applications in business analytics

📚 Key Readings

  • Core Concepts:
    • Balci, O. (1997) Principles of Simulation Model Validation, Verification, and Testing
    • Manson, S.M. (2003) Validation and verification of multi-agent models for ecosystem management
  • Machine Learning:
    • Breiman, L. (2001) Statistical modeling: The two cultures
    • James, G., Witten, D., Hastie, T. and Tibshirani, R. (2013) An introduction to statistical learning
  • Spatial Analysis:
    • Brunsdon, C., Fotheringham, A.S., Charlton, M.E. (1996) Geographically Weighted Regression
    • Tobler, W.R. (1970) A computer movie simulating urban growth in the Detroit region

📝 Assessment

The module is assessed through a 100% coursework assignment consisting of:

  • Project Report (2500 words):
    1. Construct a predictive Gradient Boosting Machine (GBM) model of house prices using the Boston Housing dataset
    2. Write up the methodology, results, and critical evaluation

Report Structure:

  1. Introduction: Problem description, study aims, advantages of GBM approach
  2. Methods: Data description, pre-processing, variable selection
  3. Results: GBM model tuning, evaluation, and application
  4. Discussion: Critical evaluation of model and results, limitations, future work

Evaluation Criteria: Understanding of task, structure, clarity/depth of argument, breadth of content, reading/referencing, writing style/presentation.

🔍 Assignment Preparation Guide

Key Steps

  1. Split the Boston Housing data into training and validation subsets
  2. Rescale the data to z-scores
  3. Create and tune a GBM model with the training subset
  4. Evaluate the model using the validation subset
  5. Analyze variable importance and model fit
  6. Critically discuss the results and methodology

Technical Requirements

  • Use R with packages: caret, gbm, tidyverse
  • Implement cross-validation for model tuning
  • Evaluate model performance using RMSE, Rsquared, and MAE

📝 Mock Assignment Brief

Task: Gradient Boosting Machine Analysis

1. Data Preparation

Using the BostonHousing dataset from the mlbench package:

 

Assessment Criteria:
  • Appropriate data splitting method (40%)
  • Justification of approach (30%)
  • Consideration of data distributions (30%)

2. Model Tuning

When creating your GBM model:

 

Assessment Criteria:
  • Identification of key parameters (30%)
  • Tuning strategy (40%)
  • Evaluation metrics (30%)

3. Critical Evaluation

After building your model:

 

Assessment Criteria:
  • Identification of limitations (40%)
  • Suggestions for improvement (40%)
  • Reference to literature (20%)