Use IBM AutoAI

1. From the project Coffee_Sales_Prediction, select New asset > Work with models, then choose Build machine learning models automatically (AutoAI). service.

2. Define the name for the AutoAI Experiment as AutoAI_coffee_sales , then select the appropriate environment. Click Create.

3. Add the coffee sales dataset, select Select data from project > Data asset > COFFEE_SALES_DATASET_shaped_2.csv > Select asset.

4. In Configure AutoAI experiment > Configure details > Create a time series analysis, select No.

5. In Configure details > What do you want to predict?, select the column SALES_CLASS. AutoAI predicts the best-suited prediction type from an initial analysis of the selected column.

6. (Optional) Select Experiment settings to customize you experiment. Customize the Prediction settings by adjusting the default values for prediction type, optimized metric, algorithm, and so on.

   • In Data source, select the training data for the model. This example uses Training data split: 90% and Holdout data split: 10%. Change other settings as required.

   • In Runtime, change the running environment to match your model training needs.

   • Select Save settings.

7. Select Run experiment.

After the experiment has run, the Pipeline leaderboard provides detailed information about the top-performing algorithm based on accuracy results. You can also view the enhancement applied to boost the pipeline's accuracy, specialization, and rank according to accuracy and build time.

1. Click the pipeline to learn more information about the different types of Evaluation used to build the pipeline, such as Model evaluation, Confusion matrix, Precision recall, Threshold.

2. Choose the best performing pipeline, Pipeline 9, select Save as. There are two options:

   • Save it as a Watson Machine Learning Model that you can test with new data, deploy to make predictions, and track its lineage activity.

   • Save it as a Notebook to view the code that created the model or interact with the model using programming language.

1. Select Model, provide a name for the model, click Create.

2. Your new model is located in the project Coffee_Sales_Prediction > Assets > Assets types > Models > AutoAI_coffee_sales - P9 LGBM Classifier - Model.

3. Select the model to view the Input Schema.

4. To deploy the model, select the Promote to space.

5. In Target space, select Create a new deployment space for the model, and provide the name as AutoAI_Deployment_Space. Click Create and select Promote.

6. Select Navigation Menu > Deployments > Spaces > AutoAI_Deployment_Space.

7. In Assets, select the model AutoAI_coffee_sales - P9 LGBM Classifier - Model

8. Select New deployment and provide the name for the deployment. Click Create. Wait for the initialization to complete.

9. To predict the daily coffee sales, select the deployment to test the model by providing input for the Test phase.

10. Our model predicts that the coffee sale will be in the medium range 2 on that particular day, with an accuracy range of 80% to 100%.

1. Select Notebook, specify the Name and Runtime environment. Click Create.

2. The notebook is located in the project Coffee_Sales_Prediction > Assets > Assets types > Notebooks > AutoAI_coffee_sales - P9 LGBM Classifier - Notebook. The notebook consists of the code used to build the model, you can review and use the code for future enhancements.