1. This article describes how to make your first AutoML model after you have gone through the data prerequisites and have successfully uploaded your dataset onto the platform. From here on, you are just a few clicks away from making meaningful predictions on your data

  2. After you upload your data the main page on the platform that you see is the predict page, the “binoculars” icon on the left panel right below the “+” upload button, with a Preview of the latest dataset that you have uploaded

  3. The “Source” button dropdown gives you the option to choose a different dataset from all the datasets that you have in your Obviously AI account. The Table gives you the name of your dataset when you upload it. You can choose the desired prediction column from the dropdown of “Pick a column to predict”. The platform chooses the last column in the dataset as the prediction column by default

  4. Next, you just click on “Go”, then you are taken to the Advanced View page (OFF by default), check for your prediction column here. You can toggle the Advanced View On and go through all the columns of the dataset as well, making sure that the ID, Prediction and Columns to Use columns have indeed the correct columns. You are now ready to make predictions on your data. Just click “Start Predicting” and off you go

  5. After you click on “Start Predicting” button, your prediction report will be ready in less than 2 minutes. The report name is generated automatically by the platform. You have the option to edit it just by clicking on the report name. So, just like that, you have created your first AutoML model

  6. The “Overview” tab on the report page, gives a quick glance into the model performance - “Great” in this case. Since this is a classification task we show the accuracy of the trained model (96.12% here) (100 being the ideal accuracy)

  7. You will also find the Drivers section here, that gives an overview of all the feature columns used for building the model and the impact of each feature column on the prediction column. As evident, the feature columns are ranked accordingly, starting with the highest impact feature to the lowest impact one. For example, in this case, for predicting “severity” of accidents, we have the accident_type feature as the most impactful one with 15% impact

  8. You can move on to the next tabs from here and also book a data science call. Moreover, you can directly go to the predictions page by clicking on the “Make Predictions” tab

  9. Right across the top corner we have 4 icons -

    1. First is for auto updating the report in case your original dataset resides on a database (currently) and the dataset has been updated with new data. Using this option you will be able to automatically re-train the model and generate a new report with the updated data

    2. Second icon is for Sync, again something particular to database tables currently, where you can manually sync the report with the updated data

    3. Third is the clone option. This is useful when you want to make changes to the columns used for prediction. Clicking on this icon opens up the advanced view and you can drag and drop columns and run predictions again. This gives the flexibility of re-training a model with a different combination of feature columns. In case you choose the clone option, a new report will be created

Did this answer your question?