MIE363H1: Operations and Supply Chain Management
Project 1: Forecasting (Winter 2023)
Project Specifications
Introduction
In this group project, you will develop forecasting models for a widely popular and lucrative use
case: predicting future stock prices. By doing so, you will gain hands-on experience in developing
and tuning forecasting models, handling real-world data, and technical report writing.
Contact Points
• For questions or issues involving methodology, data analysis, or report writing, please utilize
the Discussion feature on Quercus as others may have similar questions.
• If you have questions of a sensitive nature (e.g., personnel issues), please send an email to Ben
(ben@mie.utoronto.ca) with the subject line beginning with [MIE363].
Group Formation
Projects will be done in groups of up to 3 members. You can work alone if you wish. No groups
of 4 are allowed (we will tell you to split into groups of 2 instead). Enrol in a group using the
People tab on Quercus.
All members need to be enrolled in a project group for a submission to be considered complete;
this applies even if you are working alone. Anyone not in a Quercus group when the project is
submitted or due will be subject to mark deductions.
Stock Selection
Pick one publicly traded stock, adhering to all of the following rules:
1. The stock is traded on a North American stock exchange (e.g., NYSE, TSX, NASDAQ)
2. The stock has been trading on or before January 1, 2022
3. The stock has continuously maintained a price of at least $1.00 since January 1, 2022
4. Each group must pick a different stock:
• Declare your group’s chosen stock using the pinned Quercus discussion post
• Stock selection is on a first-come-first-serve basis
• As a courtesy to other groups, please refrain from changing your group’s selected stock
once you have made the post on Quercus
Questions regarding whether a stock is appropriate should be asked on Quercus.
Acquiring Data
Stock price data can be downloaded from Yahoo Finance (https://ca.finance.yahoo.com/) by
searching for a stock in the search bar, clicking on the “Historical Prices” tab, selecting the time
period, and then click “Download”. In this project, you are only required to analyze and predict
the stock’s closing price (i.e., the stock price at the end of the day’s trading hours), but you can
use the other included data in your analyses if you wish.
Historical Price Analysis
Conduct data analysis on the stock’s closing price between January 1, 2022 and December 31,
2022 inclusive. At minimum, you should aim to answer the following questions:
• Are there any trends and/or patterns that can be identified? Are they statistically significant?
• Were there major upward or downward spikes in the closing price? Are they related to external
events, and if so, what events were they? Use external research to justify these claims.
• Does the stock price exhibit seasonality? Is it statistically significant?
o Do not use the Dickey-Fuller test or its augmented form.
Through your data analysis, determine what type of forecasting methods would be appropriate for
use in forecasting future closing prices for the selected stock.
Future Price Forecasting
Use two appropriate methods (based on your data analysis) to forecast the stock’s closing price for
the first 10 days of January 2023. You must select one method from List A and one method from
List B (i.e., both methods cannot be from the same list).
List A List B
• Moving average
• Weighted moving average
• Exponential smoothing
• Double moving average
• Double exponential smoothing
/ Holt’s method
• Triple exponential smoothing
/ Winter’s method
• Linear or nonlinear regression^
• ARIMA or any of its other forms
(e.g., ARMA, SARIMA, SARIMAX)
• Any other prediction model, as long as
model assumptions are met^
Notes:
• For methods in List A, you may add on methods to handle seasonality (e.g., naïve method,
decomposition via moving averages) if deemed necessary.
• For methods denoted with a ^ sign, you must include at least one variable that is not either a
unit of time or a stock price (i.e., some external data must be collected and fitted to the model).
If your model contains parameters, attempt some sensitivity analysis by adjusting parameter values,
discuss why the parameters selected for prediction are optimal, and the advantages/disadvantages
of using other parameter values.
Now take the actual closing prices from the first 10 days of January 2023 and compare them against
your predicted closing prices from both your forecasting methods. Use a forecast error metric that
you believe is appropriate and justify your choice. At minimum, you should aim to answer the
following questions:
• Which forecasting method performed better? Why was it better?
• Were there external events that impacted the accuracy of your forecasts?
• Which model would you recommend to a potential investor? Would you recommend investing
in this stock at all given what you know now?
• What improvements would you make to either model? How would they improve your forecasts
(e.g., higher accuracy, higher precision)?
Deliverable
Upload 2 files to Quercus as part of your submission:
1. A PDF report with the following requirements:
• Cover page, with team members’ names and student numbers
• Report body: 8 pages maximum, double spaced (does not include cover page or references)
o You can define subsections however you like
• Formatting requirements:
o Use a standard referencing style (e.g., IEEE); no appendices allowed
o 12-point font, 1-inch margins
o Figures and tables (if any) need to be captioned
2. A zip file that contains all relevant files used in making and evaluating your forecasts such as
spreadsheets, data, code, etc.
Evaluation Criteria
You will be marked based on the following 3 criteria:
• Historical Price Analysis (30%):
o Patterns, trends, and seasonality analyses are conducted, with discussion supplemented
by statistical analysis and/or external research
o Identification of appropriate forecasting models that can be used
• Future Price Forecasting (50%)
o Appropriate and justified forecasting models developed, with sensitivity analysis
conducted as appropriate
o Comparison against actual stock prices with discussion on factors that affected forecast
accuracy
o Discussion on how to improve forecasting models for future stock price predictions
• Formatting (20%)
o Report written professionally and is easy to read
o Formatting and upload requirements are followed
Disclaimer
Our goal with this project is for you to get experience with selecting and developing forecasting
models, and stock price prediction is an inherently difficult problem. This means:
• Your mark will not be affected by the values of predicted closing prices and forecast errors.
• Although more sophisticated stock price prediction models exist, you are not required to use
them, nor will using them necessarily lead to a better mark. You can however investigate these
methods and use them if you are interested.
Additional Resources
We encourage you to look at reputable external resources (avoid blogs that don’t contain references)
to support any claims that you make regarding stock prices. Below are two types of resources that
are commonly used:
• Market reports: these are analyses conducted by market research companies to assess the
current state of particular industries. Although these reports typically cost thousands of dollars,
UofT provides access to some reports for free:
(https://guides.library.utoronto.ca/c.php?g=250479&p=1670803)
• Academic papers: these are research articles published in an academic, peer-reviewed journal.
The easiest way to find articles is typically to search for keywords on Google Scholar:
(https://scholar.google.com). If the article is behind a paywall, try searching for the article title
on the UofT Libraries site: (https://onesearch.library.utoronto.ca/)