Icon

Auto-SARIMA Exercise

SARIMA Temperature Forecasting

This workflow demonstrates how the SARIMA components can be used to generate forecasts. In this case for hourly temperature data.

Building a SARIMA Model1) Run the Table Reader node to the left.The data here is hourly temperature data for LA2) Attach and run the Timestamp Allignment component.Configure the component to hourly data, and check the box to replace the data and time column. This component will verify that our table contains data for every hour, and if not,insert a missing value.3) Attach and run the Numeric Outliers node.We include the numeric outlier node here to filter out any eronious sensor readings. We know the temperature in LA wasn't over 200F for example! Select an appropraitly large kvalue and check the box to replace outliers with missing values.4) Attach and run the Missing Values node.Now attach the missing values node and configure it to impute the missing values in our temperature column, lets use linear interpolation since temperature should be a continuousfeature.5) Connect and run the Auto-SARIMA Component.When configuring this component select the temperature as our forecast column and choose how many forecasts to generate. 72 represents 3 days and is a nice place to start. Thebackwards length setting only effects the visualization, set it to something similar to your forecast length. View the visualization.Optional 6) Use more of the components from the Time Series Analysis group.* Go to your knime explorer to the top left and connect to the example server. Double click to connect.* Open the first subfolder, 00_Components - then Time Series* Try using the inspect seasonality componet to view the ACF and PACF plots of the time series* Try using the decompose signal component to automaticaly seperate trend, seasonality, and residual* Try configuring your own SARIMA model manually with the SARIMA Learner Component* Try generating forecasts with the SARIMA Predictor component (the learner will plug into it)* Try any other components that seem intersting to you. Read TemperatureData for LA~3 weeks of data Timestamp Alignment Table Reader Auto-SARIMA Numeric Outliers Missing Value Building a SARIMA Model1) Run the Table Reader node to the left.The data here is hourly temperature data for LA2) Attach and run the Timestamp Allignment component.Configure the component to hourly data, and check the box to replace the data and time column. This component will verify that our table contains data for every hour, and if not,insert a missing value.3) Attach and run the Numeric Outliers node.We include the numeric outlier node here to filter out any eronious sensor readings. We know the temperature in LA wasn't over 200F for example! Select an appropraitly large kvalue and check the box to replace outliers with missing values.4) Attach and run the Missing Values node.Now attach the missing values node and configure it to impute the missing values in our temperature column, lets use linear interpolation since temperature should be a continuousfeature.5) Connect and run the Auto-SARIMA Component.When configuring this component select the temperature as our forecast column and choose how many forecasts to generate. 72 represents 3 days and is a nice place to start. Thebackwards length setting only effects the visualization, set it to something similar to your forecast length. View the visualization.Optional 6) Use more of the components from the Time Series Analysis group.* Go to your knime explorer to the top left and connect to the example server. Double click to connect.* Open the first subfolder, 00_Components - then Time Series* Try using the inspect seasonality componet to view the ACF and PACF plots of the time series* Try using the decompose signal component to automaticaly seperate trend, seasonality, and residual* Try configuring your own SARIMA model manually with the SARIMA Learner Component* Try generating forecasts with the SARIMA Predictor component (the learner will plug into it)* Try any other components that seem intersting to you. Read TemperatureData for LA~3 weeks of dataTimestamp Alignment Table Reader Auto-SARIMA Numeric Outliers Missing Value

Nodes

Extensions

Links

Download

To use this workflow in KNIME, download it from the below URL and open it in KNIME:

Download Workflow
Created by: corey
Created at: 2021-05-26
On NodePit since: 2022-03-22
Last update: 2024-04-26
Created with KNIME version: v4.5.1
Tags: ARIMASARIMAForecastTimeseriesOnboardingData Scientists

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!