Viewing & Analyzing Your Dataset
The Dataset View page serves as your central hub for data exploration, analysis, and preparation for machine learning. This interactive interface allows you to understand your data's characteristics, identify patterns, detect issues, and prepare for the next stages in your ML workflow.
Dataset Header & Controls
The dataset header provides essential information and powerful management controls:
Information Elements
Double-click to edit the name directly in the header
FINANCEAI-selected indicator of your dataset's domain. Proper categorization enables intelligent recommendations.
profit_marginIndicates which column is selected as your prediction target. The target selection impacts all downstream ML operations.
Action Controls
Toggles editable mode that enables column and row deletion
Permanently removes the dataset (with confirmation)
Proceeds to data preprocessing steps such as imputation, encoding, and scaling
AutoML
Warning: Any changes made while in Edit Mode (deleting rows/columns) are permanent and cannot be undone. Consider exporting a backup before making significant modifications.
Interactive Analysis Tabs
ML Clever provides seven specialized tabs for comprehensive data analysis, each offering unique insights:
Dataset Table
The primary view showing your raw data in tabular format with the following capabilities:
- View data with column headers and all rows
- In Edit Mode: Remove columns via the "X" above column headers
- In Edit Mode: Delete specific rows via the "X" at row start
- Automatic data loading as you scroll down for efficient handling of large datasets
- Sticky headers that remain visible while scrolling through data
Basic Stats
Statistical summary of each feature in your dataset:
- For numerical features: mean, median, min, max, standard deviation, and quartiles
- For categorical features: unique value counts, most frequent values, and cardinality
- Data type detection and statistics appropriate for each type
Missing Values
Visual representation of missing data patterns:
- Heatmap showing locations and patterns of missing values
- Bar chart of missing percentage by column
- Correlation between missing value patterns across columns
- Recommendations for appropriate imputation strategies
Correlation
Interactive correlation matrix visualization:
- Heatmap showing relationships between all numerical features
- Highlights high correlation pairs that may indicate redundant features
- Color-coded display from strong negative to strong positive correlations
- Tooltips showing exact correlation values on hover
Feature Distribution
Visual analysis of data distributions:
- Histograms for numerical features showing value distribution
- Bar charts for categorical features showing value frequencies
- Detection of outliers and skewness in distributions
- Split view showing distribution in relation to target variable
Feature Importance
Preliminary assessment of feature relevance:
- Bar chart ranking features by their estimated importance to the target
- Uses statistical relationships and preliminary models to assess relevance
- Helps identify potentially irrelevant features that might be candidates for removal
- Shows connection strength between each feature and your target variable
ML Readiness
Comprehensive assessment of your dataset's readiness for machine learning:
- Overall readiness score with detailed breakdown
- Checklist of potential issues that need addressing
- Personalized recommendations for preprocessing steps
- Data quality metrics and warnings about potential problems
Essential Tasks & Workflows
Dataset Management
Rename Your Dataset
Double-click on the dataset name in the header to activate inline editing. Type your new name and click anywhere outside the input to save changes.
Remove Irrelevant Columns
Click the Edit button (pencil icon) to enter Edit Mode, then click the "X" above any column header you wish to remove. Use the Correlation and Feature Importance tabs to make informed decisions about which columns might be redundant or irrelevant.
Remove Problematic Rows
With Edit Mode active, click the "X" at the beginning of any row to remove that specific data point. This is useful for eliminating obvious outliers or corrupted records that could negatively impact your model performance.
Navigate Large Datasets
Scroll through the dataset table to automatically load additional rows. ML Clever uses efficient pagination to handle large datasets without performance issues, loading new data as you scroll down.
Data Analysis Workflow
Assess Data Completeness
Start with the "Missing Values" tab to understand the extent and patterns of missing data. High percentages of missing values in certain columns may require imputation or column removal before modeling.
Explore Feature Distributions
Use the "Feature Distribution" tab to examine how values are distributed across each column. Look for heavily skewed distributions, multi-modal patterns, or extreme outliers that might require transformation or special handling.
Identify Correlated Features
Check the "Correlation" tab to discover highly correlated feature pairs (typically |r| > 0.8). Such features often contain redundant information and removing one from each highly correlated pair can simplify your model without losing predictive power.
Prioritize Important Features
Review the "Feature Importance" tab to see which variables have the strongest relationship with your target. This helps focus your attention on the most relevant features and potentially identify candidates for feature engineering.
Check ML Readiness
Finally, visit the "ML Readiness" tab for an overall assessment and specific recommendations to address any remaining issues before proceeding to preprocessing and modeling steps.
Moving Forward
After thoroughly exploring and understanding your dataset, you'll be ready to move on to preprocessing steps that prepare your data for optimal model performance:
Preprocessing Your Dataset
Click the "Preprocess Dataset" button in the header to access ML Clever’s comprehensive preprocessing suite:
- Data Imputation – Fill missing values using various statistical methods
- Feature Encoding – Convert categorical variables to numerical formats
- Feature Scaling – Normalize or standardize numerical features
- Feature Engineering – Create new features to improve model performance
- Dimension Reduction – Reduce feature space while preserving information
Pro Tips for Data Exploration
- Iterative Analysis: Data exploration is an iterative process. Revisit analysis tabs as you make changes.
- Use the Magic Button: ML Clever’s AI-powered insights can highlight issues you might miss and suggest optimal preprocessing steps.
- Data Quality First: Address missing values and outliers before advanced preprocessing.
- Documentation: Keep notes of your findings and decisions to inform your modeling strategy.