No-code ML-Lab
Test and repeat: create and run new sessions
To create a new session, follow these steps:
🟡 Step 1: Describe your session
On this screen, you define the name and a description for the session you are going to run. This will help you easily identify each attempt or variant of training you perform within the same model.
📌 Example:
Name: Energy consumption
Description: Prediction with neural networks
🟡 Step 2: Select the algorithm
In each session, you can modify the chosen algorithm when creating the training to conduct tests and compare results.
🟡 Step 3: Configure and run the training
Once the session is created, you can configure it according to the chosen model:
Configuration: define the input and output variables, the test set size, and random parameters. You can review the charts from the pre-analysis to compare the distribution of selected variables.
Hyperparameters: manually adjust the specific variables of the algorithm if you wish to fine-tune it.
Run training: click on the “Start” button to train the model.
After the process is completed, you will see the success message.
🟡 Step 4: Analyze the result
Once the training has been successfully executed, the model is ready to be evaluated. At this stage:
📊 Analyze the session metrics
By selecting “View metrics”, you access a quantitative summary of the model's performance. These metrics vary according to the type of model:
For classification, you will see accuracy, recall, F1-score, confusion matrix, among others.
For regression, mean errors (MAE, MSE), R², etc. are shown.
For clustering, metrics of cohesion and separation between groups, such as the Silhouette index, are presented.
✅ Make inferences
Click on the “Make inference” button to apply the trained model on new data. This allows you to obtain predictions with the current model. Inferences can be visualized and compared with actual data when available.
This session-based approach allows for easy iteration with different configurations and comparison of results, facilitating the tuning and validation process of the model.