Signal Characteristics
Figure: Signal Characteristics user interface
Signal characteristics TAB enables you to observe your data characteristics. You can check if the data is stationary or not and if there are any seasonality or trends present in the data. Also, you can make data stationary and remove trend or seasonality.
Objectives
By the end of this lesson, you will be
- familiar with the signal characteristics tab and
- able to learn how to use this tab for observing and extracting signal characteristics.
This tutorial assumes that you already selected a project and imported data. For more information please visit on Project and Import Data section.
User Interface Structure
Signal characteristics configuration is divided into 1. Settings 2. Signal Characteristics 3. Operation 4. Filter
Figure: Main settings
Select time feature
The time data signal has to select here. It is a single select dropdown. So user can select only one data signal here.
Select targets
You have to select the target data signal here. For example, if you have sales data that is a time frame, then sales data should be the target here.
Select other features
Other features are optional here.
Set or Edit
You have to click the set button to make the selection fixed. You can also change the selection by pressing the edit button.
Select Characteristics
There are a few options here. You have to select which characteristic of the signal you want to see. Available options are:
- Dickey-Fuller test
- Autocorrelation
- Partial Autocorrelation
- Decomposition
log or frequency
Log or frequency is optional. If you don't know the proper value, the default will be used here.
Show
Show button will display the characteristics in the plot based on the selection in the select characteristic option. Show and preview are not dependent.
Operations
After observing signal characteristics using the show button user can use various options here to make the signal usable for future use. Available operations are:
- Remove Trend
- Remove Seasonality
- Make Signal Stationary
- Resample signal
- First, Second and Third-order differencing
Preview and Apply
Preview button will do the operation but will not save the data, while Apply will give you the option to save the data in a folder in content. More information is available in here
Note
The following sections assume that you already know how to use signal characteristics tab. If not then please learn how to use signal characteristics tab from above section.
Signal Characteristics Example - Observing various characteristics of a signal
Here we will do all the steps to check signal characteristics.
Objectives
By the end of this section, you will be - able to know about various characteristics of a signal
Configurations Steps
- Select the date or time column in the select time dropdown. For example, the month signal should be selected here if the data contains monthly sales data.
- Select the target signal which has to be forecast. For example, here Shampoo_Sales data signal.
- Click on set, and selection will be locked
- Select Dickey-Fuller test/ Autocorrelation / Partial Autocorelation / Decomposition
- Keep lag/freq value blank to take the default value
- Press Show and observe the characteristics of the selected signal
Figure: Signal Characteristics
Result Discussion
You can see the results of the show button.
Signal Characteristics Example - Changing Signal Characteristics
Here we will do all the steps to change signal characteristics.
Objectives
By the end of this section, you will be - able to change a characteristic of a signal. For example, you can change a non-stationary signal into a stationary signal.
Configurations Steps
- Select the date or time column in the select time dropdown. For example, the month signal should be selected here if data contains monthly sales data.
- Select the target signal which has to be forecast. For example, here Shampoo_Sales data signal.
- Click on set, and selection will be locked
- Select Make Stationary
- Press Preview and observe the characteristics of the selected signal after the change.
Figure: Signal Characteristics
You can see the result of the operation above. At first, the Dickey-Fuller test shows it is not stationary. We made it stationary using the make stationary operation.
Understanding Signal Characteristics
Dickey-Fuller Test
When we make a model for forecasting purposes in time series analysis, we require a stationary time series for better prediction. So the first step to work on modeling is to make a time series stationary. Testing for stationarity is a frequently used activity in autoregressive modeling. We can perform various tests, and Augmented Dickey-Fuller is one of those. ADF (Augmented Dickey-Fuller) test is a statistical significance test which means the test will give results in hypothesis tests with null and alternative hypotheses. As a result, we will have a p-value from which we will need to make inferences about the time series, whether it is stationary. If p-value > 0.05 then it is non-stationary and If p-value <= 0.05 then it is stationary.
References
Dickey-Fuller Test
Check "Complete Guide To Dickey-Fuller Test In Time-Series Analysis"
Q & A
Q. What is the difference between the show and preview/apply button here?
You can see the various signal characteristics using the show button, but you can’t change any signal. Preview/Apply will change the signal based on the selected option in the operation dropdown. For example, using the dickey fuller test, the user can check a signal is stationary or not. You can make that signal stationary by using the make stationary option in operation and previewing it.
Q. When can the user use the signal characteristics tab?
A. The signal characteristics tab is significant for time series forecasting. Before using data in the time series tab for forecasting user should check and make it usable for forecasting here in the signal characteristics tab.
Q. How can I check my data is stationary or not?
A. Perform the Dickey-Fuller test and check the p-value. If it is greater than 0.05, then the time series is non-stationary.