¶ 1. Overview
Predictive analysis is a branch of advanced analysis that makes predictions about future outcomes using historical data combined with statistical modeling, data mining techniques, and machine learning. It is used to find patterns in this data in order to identify risks and opportunities.
¶ 1.1 Applications
Predictive analysis can be used for various purposes. Some of the most popular predictive analysis models are:
- Classification models: categorize data based on historical data, describing the relationships within a dataset.
- Clustering models: group data based on similar attributes.
- Time series models: use various input data at a specific time frequency.
- Other applications: fraud detection, credit risk assessment, demand forecasting, marketing optimization, predictive maintenance, etc.
¶ 2. Operation
We analyze the account history and send it to an API, which executes the algorithms and checks which is the best method to be used, returning the data to us in the form of a graph. To perform predictive analysis, a mass of previous data is required. Predictive analysis analyzes every existing data window in the application within the realized scenario.
If the selected scenario does not contain enough data, the following message will be displayed: "You need to have at least 36 periods to use predictive analysis."
Remember that the quality of predictions depends on the quality and accuracy of the presented data.
¶ 3. Utilization
User guidelines on how to select and configure the cube form for analysis, activate and interact with predictive analysis, view form data by changing values, and check for immediate interaction through the graph, as well as hiding and reusing predictive analysis.
¶ 3.1 Activation of predictive analysis
- Select the Cube Form to be analyzed;
- Select the form row to be analyzed;
- Open the tools tab at the top of the screen;
- In the cells section, select Predictive Analysis
- A panel with the prediction graph will be opened at the bottom of the screen, displaying by default a 2-year data history of the realized scenario;
- By clicking again on Predictive Analysis, in the cells section, the service is terminated.
It is noticeable that in the first use, the response time is a bit slow because the tool is analyzing the data contained in the realized dimension and selecting the algorithm to be used based on the lowest internal error rate. After the first use of the tool in the form, the next lines in which it is used will be executed instantly, as the tool analyzed the data in the first use and already selected the algorithm and parameter to be used in the form in question.
- By changing the values of the cells in the selected row, the graph is altered in real-time;
- Click apply so that the cells of the selected row are filled based on the prediction.
Predictive analysis can be used to fill out the form by clicking Apply, which will use the prediction values to fill in the form values.
Note that by clicking Apply, the forecast and prediction lines on the graph are equalized.
¶ 3.2 Minimization and Restoration
- To minimize predictive analysis, simply click on the icon
located on the right side of the tab. - To restore predictive analysis, click on the icon
located on the right side of the tab.
¶ 4. Graph Legend
-
The orange line represents the realized.
-
The blue line is the prediction, showing what was calculated according to the specific AI algorithm.
-
The red line is what we have entered in the form, a comparison between the realized and the predictive.
-
Dotted lines delimit a margin based on prediction values, so we will know the best and worst case within this scenario.
-
Prediction is a statement about what is expected to happen in the future, based on current evidence and analytical methods. It is an estimate or projection of what may occur.
-
Forecast is a statement about what is expected to happen in the future, based on current evidence and analytical methods. It is an estimate or projection of what may occur.
When changing the values of the form row, the graph is automatically adjusted.
¶ 5. Algorithms Used
Predictive analysis tests the following algorithms:
It checks the data and selects the algorithm automatically based on the lowest error rate.
¶ 5.1 Checking the algorithm being used
Hovering the mouse over the prediction legend will display the tooltip with the name of the algorithm being used.