Here to Help

ML Modeller

Estimated reading: 5 minutes 104 views

Infoveave integrates machine learning models and uses them as a tool to make predictions on the dataset. Infoveave runs and implements the ML model suited for your dataset and makes the predictions based on the selected algorithm. The machine learning models supported by Infoveave are

  • Binary Classifier to make the decision about whether the observation should be classified as positive or negative.
  • Regression is used to assess the changes in one variable and predict the changes in another variable depending on historical data.
  • Multi-Classifier is used to classify problem instances into three or more classes to analyze the dataset.
  • Clustering is used to classify objects or cases into relative clusters that show similar characteristics or patterns in a dataset.

The section covers details on 

Configure the Machine Learning Models

Steps on how to open & configure machine learning models in Infoveave

  1. To build an ML model in Infoveave,  pick the ML Modeller from Analysis ML Modeller .
    • All the ML models created if any will be displayed under the tabs My ML Models and Shared ML Models.
  1. To build a new ML model click on New. 

Define the ML Model

  1. Select the query from the drop-down list.
      1. To learn about creating Queries see here.
  2. Pick the ML Model type from binary classifier, regression, multi-classifier and clustering.
    • To advance and begin constructing the model, click Continue. The apply filters dialog box will display.

We select Multi-Classifier Model here for the explanations.

  1. Configure the required filter if the selected query contains @filter, based on time progression, dimensions and dimension items.
    • Click on the Order Date for the date progression filter
  1. To apply conditional filters click on Add Filter, select the dimension form the drop-down list.
    • For the selected dimension, pick the available conditional filters. The available conditional filters are Exactly, Not Exactly, Contains and Doesn’t Contain.  
    • Delete the applied conditional filter, using the  icon.
  • If you choose a date dimension, the condition available for you is Between, to select a date range. 
  • If the SQL query selected has no applied filters then Infoveave displays a message “Filter cannot be applied on this query”. 

Execute

  1. Click on Execute Query. With the filters applied, you will be directed to the ML Modeller window.
  2. In the ML modeller window define the model name in the Model Name String.
  3. Select the ML model type from the available options of binary classifier, regression, multi-classifier and clustering, listed in the drop-downs.

If you choose a the ML model as Clustering, always define the Cluster size. 

  1. If you need to modify the filters as per your choice, click on Filters. 
  2. To preview all the data records after the query execution click on View Table.
  1. Columns Info section elaborates columns involved in the ML modeler. By default, Infoveave defined each column as an input column with their data type.
  2. Click  to view or change the data type of that respective column.
  • Set each column as either input column or output column or ignore column.
    • Input column is the column contributing to predict the output column.
    • Output column is the column which will be predicted while analyzing the model.
    • Ignore column will be ignored while the model is built.

You must define a single output column with multiple input columns and ignore columns 

  1. Select Train-Test Ratio from drop-down list to either 70:30 or 80:20 or custom.
  2. Click on Build Model, after configuring all columns, filters, and train test ratio.
  3. Infoveave will start building models and it will display in model info with accuracy macro.
  1. After building models, select available models from the drop-down list and click on Save, a dialog box will display to confirm the Predicted Column Name.
  1. Click on Save and the model gets created and displayed in the My Model tab.

Managing ML Model

ML models can be managed with the options to share, edit, move to a folder, add a description, and delete.

  1. Share the ML model  with the team/role from the share option.
    • Choose between the options
    • To share with the users, select the “share with users” tab. Pick the user(s) with whom the ML model needs to be shared from the share dialogue box.
    • Choose the “share with role” tab to make a ML model available to a certain work role. From the dialog box, choose a role.
    • Select the users from the presented list of users.
      • If you wish to share your ML model with all, then select Share With Everyone.
    • Select the user from the available list and click  to share the ML model.
    • To remove a particular user/role from the shared list, select the user/role and click  .

    • Click Save to share your ML model with the selected users.

      • Shared personals receive a notification after the successful share.
    • To unshare a ML model, just unselect users or uncheck the Share With Everyone.
  1. Edit any of your ML model with the Edit option.
  2. To move to a folder, create a folder by clicking the New Folder option, then drag the ML model into the desired folder using the drag-and-drop feature.
    • You may also choose a folder from a drop-down menu by clicking the folder symbol on the ML model.
  1. Click “Add Description” and fill out the description details to add a description to the ML model.
    • Select the Editor Type.
    • Add necessary Tags (if required).
    • Add Description and Content.
    • Click Ok
  1. If you want to delete ML model, click on Delete .
    • Type in “delete‘ in the text box.
    • Click Yes.