Master's Project: AI-Driven
Macroeconomic Forecasting and Microeconomic Impact Analysis
Project Overview
This project challenges Master's students in Operations Research to develop an advanced machine learning model for Gross Domestic Product (GDP) forecasting. Following the successful development and validation of this macroeconomic model, students will extend their analysis to evaluate the impact of predicted GDP changes on individual companies and assess the impact of stress scenarios (as a stretch goal). This involves a multi-stage approach, combining robust data analysis, model development, and practical application to real-world scenarios.
The stretch tasks in each phase are optional. They can be ignored or answered only partially.
Project Phases
Phase 1a: Macroeconomic Data Collection and Preprocessing
Objective: To gather and prepare relevant macroeconomic data for model training.
Tasks:
● Access the macroeconomic data from FRED (https://fred.stlouisfed.org/)
● Download the data as excel files
● Identify and collect time-series data for GDP and other relevant macroeconomic indicators, explain the choice of a particular dataset
○ Real GDP
○ Personal Consumption
○ Gross Private Domestic Investment
○ CPI
○ Unemployment Rate
○ Industrial Production
○ Money Supply (M2)
○ Interest Rates
○ Oil Price
● Clean and preprocess the collected data, addressing missing values, outliers, and inconsistencies.
● Ensure data is appropriately scaled and transformed for machine learning algorithms.
Stretch Tasks:
● Use the FRED API to download the latest data
● Perform. feature engineering to create new variables that may enhance model performance.
Phase 1b: Machine Learning Model Development for GDP Forecasting
Objective: To design, implement, and evaluate a machine learning model for accurate GDP forecasting.
Tasks:
● Use Random Forests as the baseline.
● Split the dataset into training, validation, and testing sets.
● Train the chosen model using the preprocessed macroeconomic data.
● Tune hyperparameters to optimize model performance.
● Evaluate model performance using appropriate metrics (e.g., Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE)).
Stretch Tasks:
● Use an LSTM or a model of your choice for the forecasting
● Compare the performance of different models and justify the chosen model.
● Explain the results of the Random Forest model by analysing the model
Phase 2a: Data Collection and Company Selection
Objective: To gather relevant financia l and operational data for a selection of individual companies.
Tasks:
● Use the biggest 500 companies as a universe
● The following data will be provided
○ Company fundamentals
■ Return on Equity (ROE)
■ EBIT margin
■ Debt to Equity ratio
■ Market capitalization
■ Price to earnings ratio
○ Industry and sector classifications
● Use Macroeconomics data from Phase1
● Collect business-cycle or business sentiment data from FRED
● Ensure data consistency and alignment with the macroeconomic data timeline.
Phase 2b: Impact Analysis Model Development
Objective: To develop a model that quantifies the impact of forecasted GDP changes on the selected individual companies.
Tasks:
● Establish a relationship between macroeconomic factors (specifically forecasted GDP) and company-specific performance metrics (e.g., revenue growth, stock price movements, profit margins) and utilize the forecasted GDP from Phase 2 as an input to predict changes in company-specific metrics.
● Use a machine learning model for forecasting. It can be again a random forest, a LSTM or regression approach
● Consider how different sectors and business models react to economic fluctuations.
Phase 3: Stretch Goal Stress Testing
Objective: Design Stress Scenarios Tasks:
● Generate forecasts for the selected companies based on the GDP forecast.
● Conduct sensitivity analysis to understand how variations in GDP forecasts (e.g., optimistic, pessimistic scenarios) affect company performance.
● Create a heat-map of ΔROE by fi rm and scenario as well as industry and sector
These are potential stress scenarios
|
Variable
|
Baseline (today)
|
Adverse (Δ)
|
Extreme (Δ)
|
|
GDP growth
|
+2 %
|
-2 %
|
-4 %
|
|
CPI (YoY)
|
+2 %
|
+4 %
|
+6 %
|
|
Unemployment
|
4 %
|
+3 %
|
+5 %
|
|
Fed funds
|
5 %
|
+ 1 %
|
+2 %
|
|
Oil price
|
$70
|
+$30
|
+$50
|
|
10-yr yield
|
3 %
|
+0.5 %
|
+ 1.0 %
|
Phase 4: Prepare a comprehensive report detailing:
○ Methodology used for both macroeconomic and microeconomic models.
○ Model performance and validation results.
○ Specific forecasts for GDP and the selected companies.
○ Analysis of the impact of GDP changes on individual companies.
○ Limitations of the models and potential areas for future research.
○ Recommendations for businesses based on the fi ndings.
Deliverables
|
Deliverable
|
Due Date
|
Description
|
|
Project Proposal
|
Date
|
Outline of chosen methodologies, data sources, and selected
companies.
|
|
Interim Report
|
Date
|
Progress on Phase 1 and 2, initial model results, and challenges
encountered.
|
|
Final Report
|
Date
|
Comprehensive document covering all project phases, fi ndings, and analysis.
|
|
Presentation
|
Date
|
Oral presentation of the project to faculty and peers.
|
|
Code Repository
|
Date
|
All project code (e.g., Python
notebooks, scripts) submitted via a version control system (e.g.,
GitHub).
|