Sample app automatic machine learning

Using the functionality

There are two ways in which you can use the functionality:

• With VIKTOR, out-of-the-box, using our free version.
• Without VIKTOR, in that case you integrate with your own Python code.

This is a snippet of the functionality’s code. The whole code can be found and downloaded in the VIKTOR GitHub repository:

1@PlotlyView('Models', duration_guess, = 20) #for visualizing the model scores 
2def calculate_models(self, params, entity_id, **kwargs): 
3	comparison = get_model(params.dataset.data, params.dataset.target, params.choice.toggle) 
4	comparison = pd.DataFrame(comparison) 
5	cells = [comparison[col] for col in comparison.columns] 
6 
7fig = go.Figure(data=go.Table(header=dict(values=list(comparison.columns)), cells=dict(values=cells))) 
8 
9return__PlotlyResult(fig.to_json()) 

From the code snippet you can see that as little as three parameters can be used to make a machine learning model.

In the video, you can see how a dataset is uploaded, different models and analysis plots are generated and viewed, and how this data is labeled in a VIKTOR application in the demo environment.

The application contains a toggle to distinguish between classification and regression, an input field for the name/path of your .csv file so it can easily find it, a target where the column needs to be learned and distinguished, different analysis plots that can be visualized, and a table input where data ca be inserted, after which the model predicts the new property.

Apply for a demo account to get access to this and all other VIKTOR sample applications.

AI for everyone

From the video, you can see that using the machine learning application is easily done in a couple of steps.

1. In the app, the user can give the path to their .csv file, and which column they would like to predict.
2. The back-end then handles everything and conveniently shows users the result for all tested models.
3. Users can then see some plots to evaluate the model performance. The dataset is then labelled according to the model predictions so that outcomes can be compared.
4. Finally, users can see what outcomes prediction on their new data will yield.

Use the free version or apply for a demo account to try the functionality yourself!