What does it mean to predict travel?
A prediction is a careful guess about the future, based on facts from the past. For travel we want to guess two things: how crowded the road will be and how long the trip will take.
An AI (artificial intelligence) system can read much more data than a person and notice small patterns we miss.
Data: the facts the AI learns from
Data is a record of facts. For traffic it can be:
- number of cars passing a counter each hour,
- speed of phones moving along a road,
- day of the week, holiday, weather and school time,
- accidents and road work.
More data from many days gives a better picture. Data that is wrong or one-sided gives a wrong prediction.
Finding a pattern and making the prediction
A pattern is something that repeats. Traffic repeats every day: busy at about 8 am and 6 pm, calm at midday.
The simplest prediction is the average of the same hour on past days. For example, if the three past days at 8 am had 90, 96 and 94 cars, then the prediction for tomorrow is (90 + 96 + 94) ÷ 3 = 93 cars.
A real AI uses many more clues (weather, holidays, accidents) and learns how much each clue matters. This is machine learning.
From prediction to a better trip
A prediction is useful only when someone acts on it.
- Leave earlier or later to avoid the rush hour peak.
- Pick another road or take a train.
- Smart signals give more green time to the road that will be busy.
- Buses add extra trips before a big event.
Remember: a prediction can be wrong because of an accident or sudden rain. Good apps keep updating with live data. Also, they should not share where you travel without your permission.
Key formulas and definitions
- Prediction (simple) = average of the same hour on past days
- Average = sum of values ÷ number of values
- Trip time (in the 3D) = 20 × (1 + cars per hour ÷ 50) minutes
- Data → Pattern → Prediction → Decision
Worked examples
1. The cars counted at 8 am on three past days were 90, 96 and 94. Predict the count for tomorrow at 8 am.
Average = (90 + 96 + 94) ÷ 3 = 280 ÷ 3 = 93.3. So we predict about 93 cars.
2. Using trip time = 20 × (1 + cars ÷ 50), find the predicted trip time when 100 cars per hour are expected.
Cars ÷ 50 = 100 ÷ 50 = 2. Then 1 + 2 = 3. Trip = 20 × 3 = 60 minutes.
3. At 8 am the prediction is 100 cars (60 min). At 10 am it is 25 cars. How many minutes does leaving at 10 save?
At 10 am: 25 ÷ 50 = 0.5; 20 × 1.5 = 30 minutes. Savings = 60 − 30 = 30 minutes.
Common mistakes
- Thinking a prediction is always right. It is a good guess based on past patterns and can fail.
- Using only one day of data. One day may be unusual (a holiday or rain), so use many days.
- Mixing up data and pattern. Data is the raw numbers; the pattern is what repeats in them.
- Forgetting privacy. Travel data shows where people go and must be protected.