enriched dataset preview

README.md

@@ -1,5 +1,10 @@
 ---
 license: apache-2.0
+tags:
+  - tabular
+  - synthetic-data
+datasets:
+  - mostlyaiprize
 ---
 
 # MOSTLY AI Prize Dataset
@@ -41,7 +46,7 @@ This dataset consists of two CSV files used in the MOSTLY AI Prize competition:
 - File: `sequential-training.csv.gz` (1.3MB)
 - 20,000 groups
 - Each group contains 5-10 records
-- 10 data columns: 7 numeric, 3 categorical
+- 11 data columns: 7 numeric, 3 categorical + 1 group ID
 
 ### Data Format
 
@@ -99,7 +104,110 @@ sequential_data = sequential_dataset["train"]
 
 ## Dataset Schema
 
-The schema for each dataset is dynamically determined from the CSV headers. The datasets include various features relevant to the MOSTLY AI Prize competition task.
+The schema for each dataset is as follows:
+
+### Flat Dataset Schema (80 columns)
+```python
+{
+    "dog": Value("int64"),
+    "cat": Value("string"),
+    "rabbit": Value("string"),
+    "deer": Value("float32"),
+    "panda": Value("int64"),
+    "koala": Value("string"),
+    "otter": Value("string"),
+    "hedgehog": Value("float32"),
+    "squirrel": Value("int64"),
+    "dolphin": Value("int64"),
+    "penguin": Value("float32"),
+    "turtle": Value("float32"),
+    "elephant": Value("string"),
+    "giraffe": Value("int64"),
+    "lamb": Value("string"),
+    "goat": Value("string"),
+    "cow": Value("string"),
+    "horse": Value("string"),
+    "donkey": Value("string"),
+    "pony": Value("int64"),
+    "llama": Value("string"),
+    "mouse": Value("string"),
+    "hamster": Value("string"),
+    "guinea": Value("int64"),
+    "duck": Value("string"),
+    "chicken": Value("float32"),
+    "sparrow": Value("int64"),
+    "parrot": Value("int64"),
+    "finch": Value("int64"),
+    "canary": Value("int64"),
+    "bee": Value("float32"),
+    "butterfly": Value("string"),
+    "ladybug": Value("int64"),
+    "snail": Value("float32"),
+    "frog": Value("int64"),
+    "cricket": Value("int64"),
+    "tamarin": Value("string"),
+    "wallaby": Value("string"),
+    "wombat": Value("int64"),
+    "zebra": Value("int64"),
+    "flamingo": Value("float32"),
+    "peacock": Value("int64"),
+    "bat": Value("int64"),
+    "fox": Value("int64"),
+    "beaver": Value("int64"),
+    "monkey": Value("int64"),
+    "seal": Value("int64"),
+    "robin": Value("int64"),
+    "loon": Value("string"),
+    "swan": Value("int64"),
+    "goldfish": Value("int64"),
+    "minnow": Value("string"),
+    "mole": Value("float32"),
+    "shrew": Value("int64"),
+    "puffin": Value("float32"),
+    "owl": Value("int64"),
+    "bunny": Value("int64"),
+    "bear": Value("int64"),
+    "chipmunk": Value("int64"),
+    "cub": Value("string"),
+    "acorn": Value("float32"),
+    "leaf": Value("string"),
+    "cloud": Value("float32"),
+    "rainbow": Value("int64"),
+    "puddle": Value("string"),
+    "berry": Value("float32"),
+    "apple": Value("int64"),
+    "honey": Value("int64"),
+    "pumpkin": Value("string"),
+    "teddy": Value("string"),
+    "blanket": Value("string"),
+    "button": Value("string"),
+    "whistle": Value("float32"),
+    "marble": Value("int64"),
+    "wagon": Value("string"),
+    "storybook": Value("string"),
+    "candle": Value("float32"),
+    "clover": Value("float32"),
+    "bubble": Value("int64"),
+    "cookie": Value("string")
+}
+```
+
+### Sequential Dataset Schema (11 columns)
+```python
+{
+    "group_id": Value("string"),
+    "alice": Value("string"),
+    "david": Value("float32"),
+    "emily": Value("string"),
+    "jacob": Value("string"),
+    "james": Value("float32"),
+    "john": Value("string"),
+    "mike": Value("int64"),
+    "lucas": Value("float32"),
+    "mary": Value("float32"),
+    "sarah": Value("float32")
+}
+```
 
 ## Citation
 

create_visualizations.py

@@ -0,0 +1,248 @@
+#!/usr/bin/env python3
+"""
+Generate visualizations for the MOSTLY AI Prize dataset
+to enhance the Hugging Face dataset preview.
+"""
+
+import os
+import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sns
+import numpy as np
+
+# Configure visualizations
+plt.style.use('ggplot')
+sns.set(style="whitegrid")
+plt.rcParams['figure.figsize'] = (10, 6)
+plt.rcParams['figure.dpi'] = 100
+
+# Create output directory
+os.makedirs('visualizations', exist_ok=True)
+
+def load_data():
+    """Load both datasets"""
+    data_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "data")
+    
+    flat_file = os.path.join(data_dir, "flat-training.csv.gz")
+    flat_df = pd.read_csv(flat_file, compression='gzip')
+    
+    seq_file = os.path.join(data_dir, "sequential-training.csv.gz")
+    seq_df = pd.read_csv(seq_file, compression='gzip')
+    
+    return flat_df, seq_df
+
+def create_histogram(df, column, title, filename):
+    """Create a histogram for a numeric column"""
+    plt.figure(figsize=(10, 6))
+    
+    # For integer columns
+    if pd.api.types.is_integer_dtype(df[column]):
+        sns.histplot(df[column], kde=True, bins=30)
+    else:  # For float columns
+        sns.histplot(df[column], kde=True, bins=30)
+    
+    plt.title(title)
+    plt.xlabel(column)
+    plt.ylabel('Count')
+    plt.tight_layout()
+    plt.savefig(os.path.join('visualizations', filename))
+    plt.close()
+
+def create_bar_chart(df, column, title, filename):
+    """Create a bar chart for a categorical column"""
+    plt.figure(figsize=(10, 6))
+    
+    # Get value counts and limit to top 20 categories if there are many
+    value_counts = df[column].value_counts().reset_index()
+    value_counts.columns = [column, 'count']
+    
+    if len(value_counts) > 20:
+        value_counts = value_counts.head(20)
+        title += " (Top 20)"
+    
+    # Plot horizontal bar chart
+    sns.barplot(y=column, x='count', data=value_counts)
+    
+    plt.title(title)
+    plt.xlabel('Count')
+    plt.ylabel(column)
+    plt.tight_layout()
+    plt.savefig(os.path.join('visualizations', filename))
+    plt.close()
+
+def create_correlation_heatmap(df, title, filename, max_columns=20, column_subset=None):
+    """Create a correlation heatmap for numeric columns"""
+    plt.figure(figsize=(14, 12))
+    
+    # Get numeric columns
+    numeric_df = df.select_dtypes(include=['number'])
+    
+    # If specific columns are provided, use those
+    if column_subset:
+        numeric_subset = [col for col in column_subset if col in numeric_df.columns]
+        if numeric_subset:
+            numeric_df = numeric_df[numeric_subset]
+    
+    # If there are too many columns, select a subset
+    if numeric_df.shape[1] > max_columns:
+        numeric_df = numeric_df.iloc[:, :max_columns]
+    
+    # Create correlation matrix
+    corr = numeric_df.corr()
+    
+    # Create heatmap
+    mask = np.triu(np.ones_like(corr, dtype=bool))
+    sns.heatmap(corr, mask=mask, cmap="coolwarm", vmin=-1, vmax=1, 
+                annot=True, fmt=".2f", square=True, linewidths=.5)
+    
+    plt.title(title)
+    plt.tight_layout()
+    plt.savefig(os.path.join('visualizations', filename))
+    plt.close()
+
+def create_pairplot(df, columns, title, filename):
+    """Create a pairplot for selected numeric columns"""
+    plt.figure(figsize=(15, 15))
+    
+    # Create subset with the selected columns
+    subset_df = df[columns].copy()
+    
+    # Create pairplot
+    g = sns.pairplot(subset_df, diag_kind="kde", markers="o", plot_kws={"alpha": 0.6})
+    g.fig.suptitle(title, y=1.02)
+    
+    plt.tight_layout()
+    plt.savefig(os.path.join('visualizations', filename))
+    plt.close()
+
+def create_boxplot_grid(df, columns, title, filename, ncols=4):
+    """Create a grid of boxplots for selected numeric columns"""
+    # Calculate how many rows we need
+    nrows = (len(columns) + ncols - 1) // ncols
+    
+    # Create the subplots
+    fig, axes = plt.subplots(nrows=nrows, ncols=ncols, figsize=(16, 3 * nrows))
+    axes = axes.flatten()
+    
+    # Create a boxplot for each column
+    for i, col in enumerate(columns):
+        if i < len(axes):
+            if col in df.columns and pd.api.types.is_numeric_dtype(df[col]):
+                sns.boxplot(x=df[col], ax=axes[i])
+                axes[i].set_title(col)
+                axes[i].set_xlabel('')
+    
+    # Hide unused subplots
+    for i in range(len(columns), len(axes)):
+        axes[i].set_visible(False)
+    
+    plt.suptitle(title, fontsize=16)
+    plt.tight_layout()
+    plt.subplots_adjust(top=0.9)
+    plt.savefig(os.path.join('visualizations', filename))
+    plt.close()
+
+def create_categorical_distribution_grid(df, columns, title, filename, ncols=3):
+    """Create a grid of bar charts for selected categorical columns"""
+    # Calculate how many rows we need
+    nrows = (len(columns) + ncols - 1) // ncols
+    
+    # Create the subplots
+    fig, axes = plt.subplots(nrows=nrows, ncols=ncols, figsize=(16, 4 * nrows))
+    axes = axes.flatten()
+    
+    # Create a bar chart for each column
+    for i, col in enumerate(columns):
+        if i < len(axes):
+            if col in df.columns and not pd.api.types.is_numeric_dtype(df[col]):
+                # Get value counts and limit to top 10
+                value_counts = df[col].value_counts().nlargest(10)
+                value_counts.plot(kind='barh', ax=axes[i])
+                axes[i].set_title(f"{col} (Top 10 Categories)")
+                axes[i].set_xlabel('Count')
+    
+    # Hide unused subplots
+    for i in range(len(columns), len(axes)):
+        axes[i].set_visible(False)
+    
+    plt.suptitle(title, fontsize=16)
+    plt.tight_layout()
+    plt.subplots_adjust(top=0.9)
+    plt.savefig(os.path.join('visualizations', filename))
+    plt.close()
+
+def main():
+    """Main function to generate all visualizations"""
+    print("Generating visualizations for MOSTLY AI Prize dataset...")
+    
+    # Load data
+    flat_df, seq_df = load_data()
+    
+    # Create visualizations for flat dataset
+    print("Creating visualizations for flat dataset...")
+    
+    # Histograms for selected numeric columns
+    numeric_cols_flat = ['dog', 'panda', 'squirrel', 'dolphin', 'deer', 'hedgehog', 'chicken', 'bee', 'flamingo']
+    for col in numeric_cols_flat:
+        if col in flat_df.columns and pd.api.types.is_numeric_dtype(flat_df[col]):
+            create_histogram(flat_df, col, f'Distribution of {col} values', f'flat_{col}_hist.png')
+    
+    # Bar charts for selected categorical columns
+    cat_cols_flat = ['cat', 'rabbit', 'koala', 'otter', 'lamb', 'goat', 'cow', 'horse', 'llama', 'butterfly']
+    for col in cat_cols_flat:
+        if col in flat_df.columns and not pd.api.types.is_numeric_dtype(flat_df[col]):
+            create_bar_chart(flat_df, col, f'Count of {col} categories', f'flat_{col}_bar.png')
+    
+    # Grid of boxplots for numeric columns
+    create_boxplot_grid(flat_df, numeric_cols_flat, 'Boxplots of Selected Numeric Variables - Flat Dataset', 'flat_boxplots.png')
+    
+    # Grid of bar charts for categorical columns
+    create_categorical_distribution_grid(flat_df, cat_cols_flat, 'Distribution of Selected Categorical Variables - Flat Dataset', 'flat_cat_grid.png')
+    
+    # Correlation heatmap for first 20 numeric columns
+    create_correlation_heatmap(flat_df, 'Correlation Heatmap - Flat Dataset (First 20 Numeric Columns)', 'flat_correlation.png', max_columns=20)
+    
+    # Correlation heatmap for next 20 numeric columns (21-40)
+    numeric_cols_flat_next20 = flat_df.select_dtypes(include=['number']).columns[20:40].tolist()
+    if len(numeric_cols_flat_next20) > 0:
+        create_correlation_heatmap(flat_df, 'Correlation Heatmap - Flat Dataset (Numeric Columns 21-40)', 'flat_correlation_2.png', column_subset=numeric_cols_flat_next20)
+    
+    # Pairplot for selected key variables
+    key_vars = ['dog', 'deer', 'hedgehog', 'penguin']
+    if all(col in flat_df.columns for col in key_vars):
+        create_pairplot(flat_df, key_vars, 'Relationships Between Key Variables - Flat Dataset', 'flat_pairplot.png')
+    
+    # Create visualizations for sequential dataset
+    print("Creating visualizations for sequential dataset...")
+    
+    # Histograms for numeric columns
+    numeric_cols_seq = ['david', 'james', 'mike', 'lucas', 'mary', 'sarah']
+    for col in numeric_cols_seq:
+        if col in seq_df.columns and pd.api.types.is_numeric_dtype(seq_df[col]):
+            create_histogram(seq_df, col, f'Distribution of {col} values', f'seq_{col}_hist.png')
+    
+    # Bar charts for categorical columns
+    cat_cols_seq = ['alice', 'emily', 'jacob', 'john']
+    for col in cat_cols_seq:
+        if col in seq_df.columns and not pd.api.types.is_numeric_dtype(seq_df[col]):
+            create_bar_chart(seq_df, col, f'Count of {col} categories', f'seq_{col}_bar.png')
+    
+    # Grid of boxplots for numeric columns
+    create_boxplot_grid(seq_df, numeric_cols_seq, 'Boxplots of Selected Numeric Variables - Sequential Dataset', 'seq_boxplots.png')
+    
+    # Grid of bar charts for categorical columns
+    create_categorical_distribution_grid(seq_df, cat_cols_seq, 'Distribution of Selected Categorical Variables - Sequential Dataset', 'seq_cat_grid.png')
+    
+    # Correlation heatmap
+    create_correlation_heatmap(seq_df, 'Correlation Heatmap - Sequential Dataset', 'seq_correlation.png')
+    
+    # Pairplot for selected key variables
+    key_vars_seq = ['david', 'james', 'lucas', 'mary']
+    if all(col in seq_df.columns for col in key_vars_seq):
+        create_pairplot(seq_df, key_vars_seq, 'Relationships Between Key Variables - Sequential Dataset', 'seq_pairplot.png')
+    
+    print("Visualizations created successfully in the 'visualizations' directory.")
+    print("Upload these images to Hugging Face to complete the dataset preview enhancement.")
+
+if __name__ == "__main__":
+    main() 

mostlyaiprize.py

@@ -0,0 +1,259 @@
+import os
+import pandas as pd
+import datasets
+from datasets import Dataset, DatasetDict, Features, Value
+
+_CITATION = """
+@dataset{mostlyaiprize,
+  author = {MOSTLY AI},
+  title = {MOSTLY AI Prize Dataset},
+  year = {2025},
+  url = {https://www.mostlyaiprize.com/},
+}
+"""
+
+_DESCRIPTION = """
+This dataset contains the data used in the MOSTLY AI Prize competition.
+The competition focuses on synthetic data generation and evaluation.
+It contains two datasets:
+- flat-training.csv.gz: A flat (non-sequential) dataset with 100,000 records and 80 columns (60 numeric, 20 categorical)
+- sequential-training.csv.gz: A sequential dataset with 20,000 groups and 11 columns
+"""
+
+_HOMEPAGE = "https://www.mostlyaiprize.com/"
+_LICENSE = "Apache License 2.0"
+
+# Define the features for each dataset
+_FLAT_FEATURES = {
+    "dog": Value("int64"),
+    "cat": Value("string"),
+    "rabbit": Value("string"),
+    "deer": Value("float32"),
+    "panda": Value("int64"),
+    "koala": Value("string"),
+    "otter": Value("string"),
+    "hedgehog": Value("float32"),
+    "squirrel": Value("int64"),
+    "dolphin": Value("int64"),
+    "penguin": Value("float32"),
+    "turtle": Value("float32"),
+    "elephant": Value("string"),
+    "giraffe": Value("int64"),
+    "lamb": Value("string"),
+    "goat": Value("string"),
+    "cow": Value("string"),
+    "horse": Value("string"),
+    "donkey": Value("string"),
+    "pony": Value("int64"),
+    "llama": Value("string"),
+    "mouse": Value("string"),
+    "hamster": Value("string"),
+    "guinea": Value("int64"),
+    "duck": Value("string"),
+    "chicken": Value("float32"),
+    "sparrow": Value("int64"),
+    "parrot": Value("int64"),
+    "finch": Value("int64"),
+    "canary": Value("int64"),
+    "bee": Value("float32"),
+    "butterfly": Value("string"),
+    "ladybug": Value("int64"),
+    "snail": Value("float32"),
+    "frog": Value("int64"),
+    "cricket": Value("int64"),
+    "tamarin": Value("string"),
+    "wallaby": Value("string"),
+    "wombat": Value("int64"),
+    "zebra": Value("int64"),
+    "flamingo": Value("float32"),
+    "peacock": Value("int64"),
+    "bat": Value("int64"),
+    "fox": Value("int64"),
+    "beaver": Value("int64"),
+    "monkey": Value("int64"),
+    "seal": Value("int64"),
+    "robin": Value("int64"),
+    "loon": Value("string"),
+    "swan": Value("int64"),
+    "goldfish": Value("int64"),
+    "minnow": Value("string"),
+    "mole": Value("float32"),
+    "shrew": Value("int64"),
+    "puffin": Value("float32"),
+    "owl": Value("int64"),
+    "bunny": Value("int64"),
+    "bear": Value("int64"),
+    "chipmunk": Value("int64"),
+    "cub": Value("string"),
+    "acorn": Value("float32"),
+    "leaf": Value("string"),
+    "cloud": Value("float32"),
+    "rainbow": Value("int64"),
+    "puddle": Value("string"),
+    "berry": Value("float32"),
+    "apple": Value("int64"),
+    "honey": Value("int64"),
+    "pumpkin": Value("string"),
+    "teddy": Value("string"),
+    "blanket": Value("string"),
+    "button": Value("string"),
+    "whistle": Value("float32"),
+    "marble": Value("int64"),
+    "wagon": Value("string"),
+    "storybook": Value("string"),
+    "candle": Value("float32"),
+    "clover": Value("float32"),
+    "bubble": Value("int64"),
+    "cookie": Value("string")
+}
+
+_SEQUENTIAL_FEATURES = {
+    "group_id": Value("string"),
+    "alice": Value("string"),
+    "david": Value("float32"),
+    "emily": Value("string"),
+    "jacob": Value("string"),
+    "james": Value("float32"),
+    "john": Value("string"),
+    "mike": Value("int64"),
+    "lucas": Value("float32"),
+    "mary": Value("float32"),
+    "sarah": Value("float32")
+}
+
+class MostlyAIPrizeConfig(datasets.BuilderConfig):
+    """BuilderConfig for MOSTLY AI Prize dataset."""
+
+    def __init__(self, features, data_file, **kwargs):
+        """BuilderConfig for MOSTLY AI Prize.
+        Args:
+          features: Features of the dataset
+          data_file: The data file to load
+          **kwargs: keyword arguments forwarded to super.
+        """
+        super(MostlyAIPrizeConfig, self).__init__(**kwargs)
+        self.features = features
+        self.data_file = data_file
+
+class MostlyAIPrize(datasets.GeneratorBasedBuilder):
+    """MOSTLY AI Prize dataset for synthetic data generation competition."""
+
+    VERSION = datasets.Version("1.0.0")
+    
+    BUILDER_CONFIGS = [
+        MostlyAIPrizeConfig(
+            name="flat",
+            description="Flat dataset with 100,000 records and 80 columns (60 numeric, 20 categorical)",
+            features=_FLAT_FEATURES,
+            data_file="flat-training.csv.gz",
+        ),
+        MostlyAIPrizeConfig(
+            name="sequential",
+            description="Sequential dataset with 20,000 groups and 11 columns",
+            features=_SEQUENTIAL_FEATURES,
+            data_file="sequential-training.csv.gz",
+        ),
+    ]
+    
+    DEFAULT_CONFIG_NAME = "flat"
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=Features(self.config.features),
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "data")
+        data_file = os.path.join(data_dir, self.config.data_file)
+        
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "filepath": data_file,
+                },
+            ),
+        ]
+
+    def _generate_examples(self, filepath):
+        """Generate examples from the dataset file."""
+        df = pd.read_csv(filepath, compression="gzip")
+        
+        for idx, row in df.iterrows():
+            yield idx, {col: row[col] for col in self.config.features}
+            
+    # Add a method to provide dataset visualization information
+    @classmethod
+    def get_visualization_config(cls, config_name="flat"):
+        """Return configuration for dataset visualization on Hugging Face.
+        
+        This helps enhance the dataset preview with more than just a flat table.
+        """
+        if config_name == "flat":
+            return {
+                "type": "table-and-charts",
+                "charts": [
+                    {
+                        "type": "histogram",
+                        "column": "dog",
+                        "title": "Distribution of 'dog' values"
+                    },
+                    {
+                        "type": "histogram", 
+                        "column": "deer",
+                        "title": "Distribution of 'deer' values"
+                    },
+                    {
+                        "type": "histogram", 
+                        "column": "chicken",
+                        "title": "Distribution of 'chicken' values"
+                    },
+                    {
+                        "type": "bar",
+                        "column": "cat",
+                        "title": "Count of 'cat' categories"
+                    },
+                    {
+                        "type": "bar",
+                        "column": "koala",
+                        "title": "Count of 'koala' categories"
+                    }
+                ]
+            }
+        elif config_name == "sequential":
+            return {
+                "type": "table-and-charts",
+                "charts": [
+                    {
+                        "type": "histogram",
+                        "column": "mike",
+                        "title": "Distribution of 'mike' values"
+                    },
+                    {
+                        "type": "histogram",
+                        "column": "david",
+                        "title": "Distribution of 'david' values"
+                    },
+                    {
+                        "type": "histogram",
+                        "column": "james",
+                        "title": "Distribution of 'james' values"
+                    },
+                    {
+                        "type": "bar",
+                        "column": "alice",
+                        "title": "Count of 'alice' categories"
+                    },
+                    {
+                        "type": "bar",
+                        "column": "john",
+                        "title": "Count of 'john' categories"
+                    }
+                ]
+            }
+        return {"type": "table"}