Learn Forecasting In a Nutshell

 

Hello Everyone!!

Looking back, I regret being an alumnus of the mediocre Indian General MBA program (2004-2006). It left me feeling like I had received no specialized knowledge. My MBA lecturers told me that everything in MBA is a gist including forecasting and no specialized knowledge in it- I believed it and not the entire world apart from Elon Musk!! On the other hand, my MSc yielded much better results through its modules, making me an unprecedented start-up modeler and optimizer in 2024.

During COVID-19, while stuck without a job, I began forecasting the pandemic to see when I could get out of the rut and find a new job. What started as trial and error felt like blasphemy, but I gained valuable skills in constructing forecasts, interpreting results, and adjusting forecast errors with means. 

Trial & Error Method (View/Download Excel here)

Methods Utilized:

1. Single Exponential Forecast
2. Double Exponential Forecast
3. Single Linear Forecast
4. Multi-Linear Forecast
5. Least Squares Method Forecast
6. Trend Forecast
7. Monte Carlo Forecast

Procedure:

1. Constructed various forecast models and created a dashboard for displaying all results.
2. Adjusted forecast results (+/-) using Simple Average, Moving Average, and Exponential Average. Results were accurate but developed a paradox similar to the Peltzman Effect. Errors existed on a weekly basis (+/-), while forecasts were otherwise accurate within a half-day Error.
3. Used Poisson, Normal, and Geometric distributions to validate result accuracy.
4. Compared the day-before Mean adjusted forecast result with the next day's actual daily COVID-19 vaccination data. 

The delightful paradox of forecasting: I tried to mitigate the error by subtracting it with the mean, yet the error persisted once in a week, showcasing some fascinating phenomena along the way (Let's say the forecast predictor is 1000, and the mean error is 10. So, the predicted range would be 1000 +/- 10, giving us 990 to 1010. However, the next day's actual result turned out to be exactly 1000). It’s like the universe throwing a curveball just when you think you've got it all figured out. Real-life forecasting in action—unpredictable yet wonderfully intriguing!!

Outcome:

• Identified a formula and application error, leading to the abandonment of the methods & deleted off 2 other forecast methods form the sheets.

The ability to laugh at our own missteps is my rare superpower. Every great odyssey needs a few comedic interludes to make the journey worthwhile. So, feel free to laugh at my failure because I had a good laugh at it myself! How else would we ever learn to laugh in the face of failure?😄

=======================================================================

Taking a few days off to regroup, I started again with renewed vigor, ready to conquer forecasting challenges by researching formulas out of the books. It's my winding path to discovery and growth, much like any great odyssey.

Success Method (View/Download Excel here)

Methods Utilized:

1. Single Exponential Forecast
2. Single Exponential Growth Forecast
3. Double Exponential Forecast
4. Trend Regression Forecast
5. Exponential Decay Forecast
6. Exponential Growth Forecast
7. Unknown Method (For fun, I deliberately upset a formula to predict rare exponential decay)

Procedure:

• Same as above in the Trial & Error Method and introduced Seasonality Index.

Outcome:

• Excellent results without needing to adjust forecast results using Simple Average (SA), Moving Average (MVA), and Exponential Averages (EA).
• Note: Took 9 hours as daily COVID-19 vaccination time.

You can be a forecaster yourself! And hey, constructive criticism is always a good companion on the journey to accuracy. 😉

Find my older forecast experiments Here (I only got 2K views for it though 😐)

Copyrighted to ©Yasaswi Gomes

Disclaimer: The content provided in this publication is protected by copyright law. Any unauthorized use, reproduction, or distribution for personal gain apart from educational & professional gain is strictly prohibited. Please respect intellectual property rights and seek appropriate permissions if you intend to utilize this material beyond personal use.