Scribd
Upload
9 views7 pages

02 Forecasting Models

The document discusses time series forecasting models, highlighting simple models that predict future values based on past data, as well as special models like Exponential Smoothing, ARIMA, and Prophet that can utilize raw time series data. It also emphasizes the need for predictors when using traditional machine learning models such as linear regression and random forests, which require feature engineering. The main focus is on generating predictors and targets for effective forecasting.

Uploaded by

sharma94.chirag
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views7 pages

02 Forecasting Models

The document discusses time series forecasting models, highlighting simple models that predict future values based on past data, as well as special models like Exponential Smoothing, ARIMA, and Prophet that can utilize raw time series data. It also emphasizes the need for predictors when using traditional machine learning models such as linear regression and random forests, which require feature engineering. The main focus is on generating predictors and targets for effective forecasting.

Uploaded by

sharma94.chirag
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Forecasting

models
Machine Learning
Algorithms

1
Time Series Forecasting

Time Sales

30/03/20 200

31/03/20 220
• We don’t have predictors
01/04/20 230
• We need to use past data
02/04/20
Time 235
Sales
03/04/20 ?
04/04/20 ?

2
Time Series Forecasting

Time Sales
Simple models:
30/03/20 200

31/03/20 220
• Predict the last value
01/04/20 230

02/04/20
Time 235 • Predict the mean of the last x values
Sales
03/04/20 ?
(moving average)
04/04/20 ?

3
Time Series Forecasting

Time Sales Special models:


30/03/20 200

31/03/20 220
• Exponential Smoothing: weighted average of last values
01/04/20 230
• ARIMA - Autoregression: forecast with a linear
02/04/20
Time 235
Sales
combination of past values of the target.
03/04/20 ?
04/04/20
• Prophet
?
• Neuronal networks - RNN

• These models can take the “raw” time series as input.


• Neural Networks can also take additional features.

4
Time Series Forecasting

Time Sales Off-the-shelf machine learning models


30/03/20 200

31/03/20 220
• Linear regression
01/04/20 230
• Random Forests
02/04/20
Time 235
Sales
• Gradient Boosting Machines (xgb, lightGBMs)
03/04/20 ?
04/04/20
• SVM, KNN, etc
?
• We need predictors!

• These models predict a target based of predictors.


• We need predictor variables.
• We need a suitable target.
5
Extract features
Target Predictors
Machine
Time Sales Var 1 Var 2 Var 2 Var 4 Var 5 Var 5
Learning
30/03/20 200 Model 3 30 15 50 0.2 0

31/03/20 220 3 31 16 10 0.2 0

01/04/20 230 4 1 17 0 0.19 1

02/04/20
Time 235
Sales 4 1
Var 2 2
Var 19 2
Var 54
Var 0.175
Var 0 5
Var
03/04/20 ? 3 30 15 50 0.2 0
04/04/20 ? 3 31 16 10 0.2 0

• How do we generate the predictors and the target?


 This is the main topic of the course.
6
Simple models predict future values
based of past values.

Summary Special models that can take the raw


time series, like ES, ARIMA and Prophet.

We can forecast with traditional


machine learning models plus feature
engineering.

You might also like