ML Operations


Before you can make any prediction, you need to create a new model using the dataset uploaded. When you are building your model, you can view the status of the model through the model dashboard list view.

Click to expand image

Dashboard List

Model dashboard list

The list shows the details of the model you want to build or predict, it shows the following element:

The Title of the model you want to give the model
It shows the model type of the model either prediction model type or clustering model type.
The day, month and year it was created on.
You can view the status of the model.

Creating a new model

To create a new model, click on the create new model button at the top right of the model dashboard, it takes you to where you will fill in information about the type of model you want to create either a prediction type type or clustering type.

Prediction type

In choosing the prediction type, you must fill in the model name, select the dataset and also select the target variable you want to predict and then create the model.

Click to expand image

Prediction Type
Quick tip

When you choose the prediction type of model, you should understand the drivers of your business and predict what would happen next: cross-selling, churn, fraud, likeliness to click or to purchase.

Clustering type

In using the clustering type, you fill in the information about the model name and select the dataset you want to predict and then create the model.

Quick tip

You must look for hidden patterns and discover groups of people sharing the same behaviour. Discover how they are using your product or services.

Viewing a model

When creating a model, you can view the models that are running, failed, or completed. When viewing the model you are building, there are feature elements to note which include:

Model name: this is the name of the model your datasets are running on.

Metrics: it shows the optimization and evaluation performance of the model in form of decimal numbers.

Status: this is the status of the model either it is running, failed or completed.

Click to expand image

View Model

When you click on the model which you created under the view model dashboard, there will be a drop-down of every element of that particular model. These include:

Click to expand image


Metrics: it shows how well the model is performing or have performed, it explains the Variance Score and Mean Absolute Error(MAE) of the models.

Confusion matrix: it shows the total percentage(%) of actual classes of prediction of the models, that is the shipped, in progress, resolved, disputed, on hold, cancelled.

Roc curve: it shows the ROC_AUC for the prediction classes which include shipped, unshipped and flagged. Select the class you intend to see the AUC.number.

Variable importance: it shows the variable representation in form of chart.

Algorithms: The algorithms show the element of the trained data used for the model, it shows the numbers of rows and columns before and after processing the model, it shows the matrix type, estimated memory usage of the model.

Features: These are the elements selected for the model development.

Train and validation: Train is when the model sees and learns from the datasets and validation is the process of verifying that the models are providing a satisfactory fit to their data, in line with both qualitative and quantitative objectives. Validation is for evaluating a model and also fine-tune the model hyperparameters, the training and validation of the models show the policy adopted for the model, the sampling method used, the partitioning of the data, the records limits, the splitting mode and the train ratio of the models. The train and validation also show the list element of train and test sets of datasets used for the model, that is date and time the data was generated, the numbers of rows of the train set and the numbers of rows of the test sets.

Prediction:under the prediction section of the model, there are elements to take note of which includes:

Test prediction: In order to preview how a model returns prediction, score up to 100MB to preview how the model returns prediction. To run an external test when generating prediction, upload a dataset that includes the target. Optionally include up to five features from your dataset with the downloaded predictions.

Prediction datasets: To upload datasets or import a file, click on the import form to select where you want to upload the dataset from either from a local file or from a data source and after choosing the dataset, click on compute prediction.

Click to expand image

Test Prediction

From the local file: To choose a datasets from a local file, click on the import from and click on the local file and select from where you want to upload the dataset from on the system and then upload.

From a data source: To upload from a dataset from a data source for prediction, click on import from, select the data source option, click the dataset you want to upload.

Did you find what you were looking for?
Yes No