Hello :) Today is Day 7!
A quick summary of today:
- WON/USD exchange rate prediction model
- Seoul Bike Rental Demand - Linear Regression Model and Seasonal Features
Following yesterday’s Kaggle time series course, I decided to apply what I learned. The links go to the Kaggle notebooks for each model. Below I will try to provide an overview of both.
WON/USD exchange rate prediction model
Data
I used simple interpolation to deal with missing data points
Feature engineering
lag_1 - made 1 day behind lag feature
Model evaluation R-squared: 0.9941 Mean Squared Error: 63.3640 Mean Absolute Error: 4.6090
Homoscedasticity check
The model results seem good, but I know that the used features are not that good because I only use lag_1 (I just applied what I learned yesterday)
Seoul Bike Rental Demand
Looked to do time series forecast using Seoul Bicycle Demand Data, which includes daily and hourly information and weather information on bicycles rented in Seoul. The goal was to construct a basic linear regression model that integrates seasonal features and holiday information.
Some data viz
Checked for a trend (seems like there is not)
Seasonality check
Seems there is annual, and no daily.
After adding seasonality features, the model results:
Also made a 2nd model, that includes holiday data:
Results
That is all for today!
See you tomorrow :)