Every month, I travel twice to a western Mumbai suburb for a meeting. Usually, the time starts at 9.30 am. I live in Pen and, therefore, really need to plan my travel. The planning is based on several factors:
1. Construction work on Goa Highway.
2. Season - Monsoon and Non-Monsoon and days like Diwali, Holi, Christmas, etc.
3. Time of the day (Early Morning, Mid-morning, Afternoon, and Evening).
4. Day of the week.
I love to capture and analyze data. Holt-Winters Model is a time series forecasting method incorporating trend, seasonality, and level components into its predictions. The model is based on three factors:
a) Level Component (α - Alpha) - In this case, the average commute time.
b) Trend Component (β - Beta) -The rate of change in commute time over time due to factors like traffic patterns or road construction.
c) Seasonal Component (γ - Gamma) - The seasonal component captures periodic fluctuations or patterns in the data i.e. weekly variations in commute time, such as longer commute times on Mondays compared to Fridays.
Using historical data and the Holt-Winters equations, forecast the commute time for future days. The forecasted commute time for a specific day is the sum of the level, trend, and seasonal components.
So, today, I started at 6.15 am. Google Maps predicted the time to reach Malad as 7.52 am. Holt-Winters model predicted a time of 8.30 am. I reached the client's office at 8:05. Almost 1:30 hour before the meeting time! But, in the past, I had once started between 7.30 and 8 am after checking Google Maps and the values predicted by the model. On that day, I reached Malad close to noon!
Western Express makes the model sweat!
But, I will not give up. Someday, I will find the accurate α, β, and γ values for my commute to Malad. 😁
Subodh
Comentarios