Icon

AI PEPC ARIMA Forecasting - Crude Oil Price WTI

<p>The <strong>ARIMA</strong> (AutoRegressive Integrated Moving Average) model is a powerful statistical method used for analyzing and forecasting time series data. It handles standard patterns in time series and offers a simple yet effective approach to forecasting.</p><p>ARIMA combines three components:</p><ul><li><p><strong>AR (Autoregression):</strong> Uses past values to predict future ones.</p></li><li><p><strong>I (Integrated):</strong> Applies differencing to make the data stationary by removing trends.</p></li><li><p><strong>MA (Moving Average):</strong> Models the relationship between observations and past forecast errors.</p></li></ul><p>The model is typically expressed as <strong>ARIMA(p, d, q)</strong> where:</p><ul><li><p>p is the number of lag observations,</p></li><li><p>d is the degree of differencing,</p></li><li><p>q is the order of the moving average.</p></li></ul><p>ARIMA builds a linear regression model using these components. Setting any parameter to zero simplifies the model to AR, MA, or ARMA. The model assumes the data follows an ARIMA process, making it essential to check whether the data and residuals meet this assumption.</p><p></p><p>This workflow demonstrates the simple use of ARIMA for forecasting on time series data. This workflow includes:</p><ul><li><p>data preparation (data import and preproccessing)</p></li><li><p>data analysis (stationary and auto correlation test)</p></li><li><p>build the model</p></li><li><p>evaluate the model</p></li><li><p>visualization</p></li></ul><p>The simple dataset used consists of only 2 columns, namely <strong>dates</strong> and <strong>y</strong> values.</p>

URL: Building a time series analysis application https://www.knime.com/blog/building-a-time-series-analysis-application
URL: What is ARIMA Model https://www.ibm.com/id-id/think/topics/arima-model

Data Import
Data Preprocessing
Build the model
Visualization
  • plot original data (y)

  • plot y data for model testing

  • comparation plot: y and predicted y

Data Analysis

ARIMA - Simple Time Series Forecasting

demonstrate simple use ARIMA for forcasting on time series data

this workflow includes:

  • data preparation (data import and preprocessing)

  • data analysis (stationary and autocorrelation test)

  • build the model

  • evaluate the model

  • visualization


test model
ARIMA Predictor
rename column: dates
Column Renamer
train = 518 first datetest = 7 last date
Table Partitioner
Statistics View
Introduce missing days
Timestamp Alignment
plot y (original values) for testing only
Line Plot (JavaScript) (legacy)
Excel Reader
sort dates ascending
Sorter
Evaluate model: calculate score metric(R2, MAE, MSE, RMSE, MSD)
Numeric Scorer
show ACF and PACFuseful to check: data relationship to previous data in time time series
Autocorrelation Plot (Labs)
join y and y prediction(y(fitted values))
Joiner
String to Date&Time
compare: y (original) with y(fitted values) as prediction
Line Plot (JavaScript) (legacy)
train modeluse ARIMA with input parameter (p,d,q)
ARIMA Learner
check data stationarity using ADF Test
ADF Test (python)
plot all y data (original)
Line Plot (JavaScript) (legacy)
set y missing value with 0
Missing Value
result:data is time series stationer(p-value < 0.05)notes:this result important to decide d parameter on ARIMA
Table View

Nodes

Extensions

Links