Creating a Predictor¶
If you have not read our Predictor Overview guide, we highly recommend you read that before creating a predictor. It will help you understand how to structure your data and what's going on under the hood.
Step 1. Choosing Your Objective Column¶
When you create your first predictor on Apteo, you'll be brought to a screen that prompts you to select the column in your dataset that contains the key metric that you want to estimate.
Once you select your column, you can click the "PREDICT" button, which will begin the process of training and evaluating multiple models. By default, the system will use all of the columns in the same table as your key metric as your "training data" (or the data that is used to estimate your key metric).
Step 2. Wait For Training To Finish and Name Your Model¶
The system will then begin to train and evaluate various machine learning models. Depending on the size of your data, this process could take anywhere from a few minutes to half an hour. During this time you can name your model using the input box at the top left of the screen.
Step 3. Predict¶
In order to use a trained predictor to create a new prediction, you need to provide it with values for all of the same input fields that were used to train that predictor. The system allows you to do so in three different ways:
- Creating a single prediction by filling out a form
- Creating multiple predictions at once by uploading a CSV file that has all of the data that the predictor needs
- Using the prediction API (available for workspaces on a Professional or Enterprise plan)
To create a single prediction using a trained predictor, you can click on that predictor from the "Predict" link in the left side navigation bar. You'll be shown a form which will allow you to enter values for the fields that were used to train your model. The goal is to use known values to predict a new value.
If you'd like to create more than one prediction a time, you can alternatively use the Batch Prediction screen. You'll need to upload a CSV that contains filled out values for each of the attributes used to train the predictor. You can then upload that CSV and the system will provide you with predictions for all of the rows in the CSV, up to a maximum of 500 predictions.
Predictions via API¶
For workspaces on a Professional or Enterprise plan, members have the option of creating predictions via an API. API predictions for every predictor are provided on the "PREDICTION API" tab in the trained predictor's results screen. The prediction API is limited to 5,000 requests per hour and 1 million requests per month.