whackthejacker/DataHubHub

1import streamlit as st2import pandas as pd3import numpy as np4import plotly.express as px5import plotly.graph_objects as go6 7def render_dataset_statistics(dataset, dataset_type):8    """9    Renders statistical analysis of the dataset.10    11    Args:12        dataset: The dataset to analyze (pandas DataFrame)13        dataset_type: The type of dataset (csv, json, etc.)14    """15    if dataset is None:16        st.warning("No dataset to analyze.")17        return18    19    st.markdown("<h3>Dataset Statistics</h3>", unsafe_allow_html=True)20    21    # Tabs for different kinds of statistics22    tab1, tab2, tab3 = st.tabs(["Summary Statistics", "Distribution Analysis", "Correlation Analysis"])23    24    with tab1:25        # Summary statistics26        st.markdown("### Summary Statistics")27        28        # Filter only numeric columns for statistics29        numeric_cols = dataset.select_dtypes(include=[np.number]).columns.tolist()30        31        if numeric_cols:32            # Display summary statistics33            st.dataframe(dataset[numeric_cols].describe().T.style.highlight_max(axis=1, color='#FFD21E'), use_container_width=True)34            35            # Top values for categorical columns36            categorical_cols = dataset.select_dtypes(exclude=[np.number]).columns.tolist()37            if categorical_cols:38                st.markdown("### Category Value Counts")39                selected_cat_col = st.selectbox("Select categorical column", categorical_cols)40                41                # Show top values and their counts42                value_counts = dataset[selected_cat_col].value_counts().head(10)43                fig = px.bar(44                    x=value_counts.index, 45                    y=value_counts.values,46                    title=f"Top 10 values in {selected_cat_col}",47                    labels={"x": selected_cat_col, "y": "Count"},48                    color_discrete_sequence=["#2563EB"]49                )50                st.plotly_chart(fig, use_container_width=True)51        else:52            st.warning("No numeric columns found in the dataset.")53    54    with tab2:55        # Distribution analysis56        st.markdown("### Distribution Analysis")57        58        if numeric_cols:59            selected_num_col = st.selectbox("Select numeric column", numeric_cols)60            61            # Create distribution plot62            fig = px.histogram(63                dataset, 64                x=selected_num_col,65                title=f"Distribution of {selected_num_col}",66                marginal="box",67                color_discrete_sequence=["#FFD21E"],68                template="simple_white"69            )70            st.plotly_chart(fig, use_container_width=True)71            72            # Basic distribution stats73            col1, col2, col3, col4 = st.columns(4)74            with col1:75                st.metric("Mean", f"{dataset[selected_num_col].mean():.2f}")76            with col2:77                st.metric("Median", f"{dataset[selected_num_col].median():.2f}")78            with col3:79                st.metric("Min", f"{dataset[selected_num_col].min():.2f}")80            with col4:81                st.metric("Max", f"{dataset[selected_num_col].max():.2f}")82        else:83            st.warning("No numeric columns found in the dataset.")84    85    with tab3:86        # Correlation analysis87        st.markdown("### Correlation Analysis")88        89        if len(numeric_cols) > 1:90            # Compute correlation matrix91            corr_matrix = dataset[numeric_cols].corr()92            93            # Plot heatmap94            fig = px.imshow(95                corr_matrix,96                color_continuous_scale=["#84919A", "#FFFFFF", "#FFD21E"],97                title="Correlation Matrix",98                template="simple_white"99            )100            st.plotly_chart(fig, use_container_width=True)101            102            # Top correlated features103            st.markdown("### Top Correlated Features")104            105            # Convert correlation matrix to a long format106            corr_pairs = []107            for i in range(len(corr_matrix.columns)):108                for j in range(i+1, len(corr_matrix.columns)):109                    col1 = corr_matrix.columns[i]110                    col2 = corr_matrix.columns[j]111                    corr_value = corr_matrix.iloc[i, j]112                    corr_pairs.append((col1, col2, corr_value))113            114            # Sort by absolute correlation115            corr_pairs.sort(key=lambda x: abs(x[2]), reverse=True)116            117            # Display top 10 correlated pairs118            if corr_pairs:119                top_pairs = pd.DataFrame(corr_pairs[:10], columns=["Feature 1", "Feature 2", "Correlation"])120                st.dataframe(121                    top_pairs.style.format({122                        "Correlation": "{:.4f}"123                    }).background_gradient(subset=["Correlation"], cmap="coolwarm"),124                    use_container_width=True125                )126                127                # Scatter plot for the top correlated pair128                if corr_pairs:129                    top_pair = corr_pairs[0]130                    fig = px.scatter(131                        dataset, 132                        x=top_pair[0], 133                        y=top_pair[1],134                        title=f"Scatter plot: {top_pair[0]} vs {top_pair[1]} (Corr: {top_pair[2]:.4f})",135                        color_discrete_sequence=["#2563EB"],136                        template="simple_white"137                    )138                    fig.add_traces(139                        go.Scatter(140                            x=[None], 141                            y=[None],142                            mode='lines',143                            line=dict(color="#FFD21E", width=3),144                            name='Best Fit'145                        )146                    )147                    st.plotly_chart(fig, use_container_width=True)148        else:149            st.warning("Need at least two numeric columns for correlation analysis.")150