Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

Phase 1: Problem Formulation and Economic Context

1. Strategic Problem Definition: Select a high-impact business decision (e.g., dynamic pricing, credit risk assessment, algorithmic trading, customer retention, or supply chain forecasting). Review the peer-reviewed academic literature (economics, finance, management, and computer science) to explain the current state of research and develop a thorough theoretical framework for understating the business problem (e.g., customer retention). How may the emergence of large dataset and ML algorithms (e.g., pricing algorithms) improve the business decision and increase performance?

2. The Data Generating Process: Select a real-world dataset (from reputable sources or Canvas).

Beyond just describing the data, critically analyse the Data Generating Process (DGP). Are there inherent selection biases in how the data was collected? What variables are missing that domain experts would consider crucial?

Phase 2: Exploratory Analysis and Feature Engineering

3. Theory-Driven Feature Engineering: Explain your data cleaning and preparation steps. Crucially, construct new variables (feature engineering) motivated by economic or business theory, not just statistical correlation.

4. Exploratory Data Analysis (EDA): Use advanced data visualisation to reveal key patterns. Explain how this exploratory phase alone could alter managerial decision-making, even before an ML model is trained.

Phase 3: Algorithm Development and Business Evaluation

5. Model Development: Train at least three distinct machine learning models (e.g., Regression Trees, Gradient Boosting/XGBoost, Random Forests, or Neural Networks). If your problem involves text or images, you may incorporate NLP or deep learning techniques. Intuitively explain your ML techniques.

6. Model Comparison / Selection: Explain the model selection process using cross-validation, report your results and select your optimal model (algorithm).

7. Illustrate the model in practice by demonstrating how it can be used in practice to make predictions and support business decisions.

Phase 4: Strategic Implications, Causality, and Ethics

8. Prediction vs. Causality: Machine learning is generally predictive, not causal. Reflect on this limitation. For example, if your model predicts churn, it doesn't automatically tell you what intervention will stop it. Discuss how a firm would need to design an experiment (e.g., A/B testing, RCT) to test the algorithm's actual effectiveness in the real world.

9. Algorithmic Fairness and Regulation: Analyse your model for potential biases. Could the deployment of this model result in unfair treatment of certain demographics? How should the business navigate data privacy laws (e.g., GDPR, CCPA) and ethical considerations when deploying this AI?

10. The Human-AI Synthesis: Conclude by reflecting on the limits of automation in your chosen domain. In what specific ways must domain expertise and "human-in-the-loop" oversight be integrated with your algorithm to ensure long-term business success?

Include your code with comments in an appendix at the end of the essay. The code will not be counted towards the word limit.

Students are required to strictly and only consult top academic journals ranked 4 stars and provide the page number for any reference used in the project. References to non-peer reviewed sources, such as blogs, consultancy reports, and online websites, and failure to provide precise page number will automatically result in a mark below 40.

Word Limit: Strictly 3000 words. The 10% rule applies.

Deadline: The coursework should be submitted online via Canvas by 2 p.m. on Friday, 15 May 2026.