SimLabs ML Lab

Machine Learning Modeling & Analysis Tool

This page focuses on the upload, preview, visualization analysis, training, and prediction workflow for small to medium-sized datasets. It is suitable for educational experiments, method demonstrations, sample analysis, and rapid validation, and is not intended as a large-scale production AutoML platform.

File Upload

Supports .csv, .xlsx, .json formats. Data will be processed entirely in your local browser.

Recommended file size < 50MB

Data Overview

No data availableDisplay partial data

Exploratory Data Analysis (EDA)

Machine Learning

Current Demo: Simple Neural Network Model (Predicting Numeric Values)
Model Configuration
Hold Ctrl/Cmd to select multiple
Training Parameters
Network Architecture
Early Stopping Strategy
Training Monitor (Loss)
Model Evaluation & Prediction
Training has not started yet...

Advanced Visualization Analysis

Provides in-depth data insights and model interpretation based on the trained model.
Feature Correlation
Prediction vs Actual
Residual Analysis
Feature Importance

Feature Correlation Matrix

Displays the Pearson correlation coefficient between numeric features to help understand the relationships among them. The system will automatically detect numeric columns in the data.

Click matrix cells to view details

Click the button above to generate the correlation matrix

Requires at least two numeric columns

Interpretation Guide
Strong Positive Correlation (0.7~1.0)
Moderate Positive Correlation (0.3~0.7)
Weak Correlation (0.0~0.3)
Strong Negative Correlation (-1.0~-0.7)
Practical Tips
  • The diagonal shows the feature's relationship with itself (always 1)
  • Highly correlated features (|r| > 0.8) may be redundant
  • Strong correlation with the target variable often indicates good predictive power
  • Check negative correlations, which may reveal interesting patterns

Predicted vs Actual Values

Compare the match between model predictions and actual values. Ideally, points should fall on the diagonal line.

Model training must be completed first
Ideal Line: y = x (Predicted = Actual)
R² Score: -
Perfect Prediction: All points fall on the diagonal line
Mean Error: -

Residual Analysis

Examine the model's error distribution. Ideally, residuals should be randomly distributed around the zero line.

Residual = Actual Value - Predicted Value
Residual Statistics
Mean
-
Standard Deviation
-
Normality Test
-
Ideal Scenario:
  • Residual mean is 0
  • Residuals follow a normal distribution
  • No obvious patterns or trends

Feature Importance Analysis

Feature importance calculated based on model weights, helping understand which features contribute most to predictions.

Based on trained neural network weights
Importance Interpretation

The importance score indicates the relative contribution of a feature to the model's prediction:

  • High importance: The feature has a large impact on predictions
  • Low importance: Consider removing it from the model
  • Negative importance: Negatively correlated with the target variable
Feature Suggestions
Click "Calculate Feature Importance" to get suggestions