data-science-expert — скилл для Claude Code: анализ данных

---
name: data-science-expert
version: 1.1.0
description: >-
  Expert-level data science, analytics, visualization, and statistical modeling. Use when
  the user mentions analytics, visualization, statistics, pandas, or NumPy, or when the
  task involves Data Analysis, Machine Learning, Data Visualization, or Feature
  Engineering.
category: ai
tags: [data-science, analytics, visualization, statistics, pandas, numpy]
allowed-tools:
  - Read
  - Write
  - Edit
  - Bash(python:*)
---

# Data Science Expert

Expert guidance for data science, analytics, statistical modeling, and data visualization.

## Core Concepts

### Data Analysis

- Exploratory Data Analysis (EDA)
- Data cleaning and preprocessing
- Feature engineering
- Statistical inference
- Time series analysis
- A/B testing

### Machine Learning

- Supervised learning (classification, regression)
- Unsupervised learning (clustering, PCA)
- Model selection and validation
- Feature importance
- Hyperparameter tuning
- Ensemble methods

### Data Visualization

- Matplotlib, Seaborn, Plotly
- Statistical plots
- Interactive dashboards
- Storytelling with data
- Best practices for visualization
- Color theory and accessibility

## Data Cleaning and EDA

```python
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from typing import Dict, List

class DataCleaner:
    """Clean and preprocess data"""

    def __init__(self, df: pd.DataFrame):
        self.df = df.copy()
        self.cleaning_log = []

    def handle_missing_values(self, strategy: str = 'drop',
                             fill_value=None) -> pd.DataFrame:
        """Handle missing values"""
        missing_before = self.df.isnull().sum().sum()

        if strategy == 'drop':
            self.df = self.df.dropna()
        elif strategy == 'fill':
            if fill_value is not None:
                self.df = self.df.fillna(fill_value)
            else:
                # Fill numeric with median, categorical with mode
                for col in self.df.columns:
                    if self.df[col].dtype in ['float64', 'int64']:
                        self.df[col].fillna(self.df[col].median(), inplace=True)
                    else:
                        self.df[col].fillna(self.df[col].mode()[0], inplace=True)

        missing_after = self.df.isnull().sum().sum()
        self.cleaning_log.append(f"Missing values: {missing_before} -> {missing_after}")

        return self.df

    def remove_duplicates(self) -> pd.DataFrame:
        """Remove duplicate rows"""
        before = len(self.df)
        self.df = self.df.drop_duplicates()
        after = len(self.df)

        self.cleaning_log.append(f"Duplicates removed: {before - after}")
        return self.df

    def remove_outliers(self, columns: List[str],
                       method: str = 'iqr',
                       threshold: float = 1.5) -> pd.DataFrame:
        """Remove outliers"""
        before = len(self.df)

        for col in columns:
            if method == 'iqr':
                Q1 = self.df[col].quantile(0.25)
                Q3 = self.df[col].quantile(0.75)
                IQR = Q3 - Q1

                lower = Q1 - threshold * IQR
                upper = Q3 + threshold * IQR

                self.df = self.df[(self.df[col] >= lower) & (self.df[col] <= upper)]

            elif method == 'zscore':
                z_scores = np.abs(stats.zscore(self.df[col]))
                self.df = self.df[z_scores < threshold]

        after = len(self.df)
        self.cleaning_log.append(f"Outliers removed: {before - after}")

        return self.df

class EDA:
    """Exploratory Data Analysis"""

    def __init__(self, df: pd.DataFrame):
        self.df = df

    def summary_stats(self) -> pd.DataFrame:
        """Generate summary statistics"""
        return self.df.describe(include='all').T

    def correlation_analysis(self, method: str = 'pearson') -> pd.DataFrame:
        """Calculate correlation matrix"""
        numeric_cols = self.df.select_dtypes(include=[np.number]).columns
        return self.df[numeric_cols].corr(method=method)

    def plot_distributions(self, columns: List[str] = None):
        """Plot distributions of numeric columns"""
        if columns is None:
            columns = self.df.select_dtypes(include=[np.number]).columns

        n_cols = len(columns)
        n_rows = (n_cols + 2) // 3

        fig, axes = plt.subplots(n_rows, 3, figsize=(15, 5*n_rows))
        axes = axes.flatten()

        for idx, col in enumerate(columns):
            sns.histplot(self.df[col], kde=True, ax=axes[idx])
            axes[idx].set_title(f'Distribution of {col}')

        plt.tight_layout()
        return fig

    def plot_correlation_heatmap(self):
        """Plot correlation heatmap"""
        corr = self.correlation_analysis()

        plt.figure(figsize=(12, 10))
        sns.heatmap(corr, annot=True, fmt='.2f', cmap='coolwarm',
                   center=0, square=True, linewidths=1)
        plt.title('Correlation Heatmap')
        return plt.gcf()
```

## Feature Engineering

```python
from sklearn.preprocessing import StandardScaler, LabelEncoder, OneHotEncoder
from sklearn.feature_selection import SelectKBest, f_classif, mutual_info_classif

class FeatureEngineer:
    """Engineer features for machine learning"""

    def __init__(self, df: pd.DataFrame):
        self.df = df.copy()
        self.transformers = {}

    def create_interaction_features(self, col1: str, col2: str) -> pd.Series:
        """Create interaction features"""
        self.df[f'{col1}_x_{col2}'] = self.df[col1] * self.df[col2]
        return self.df[f'{col1}_x_{col2}']

    def create_polynomial_features(self, col: str, degree: int = 2) -> pd.DataFrame:
        """Create polynomial features"""
        for d in range(2, degree + 1):
            self.df[f'{col}_pow_{d}'] = self.df[col] ** d
        return self.df

    def bin_numeric_feature(self, col: str, n_bins: int = 5,
                           strategy: str = 'quantile') -> pd.Series:
        """Bin numeric features"""
        self.df[f'{col}_binned'] = pd.qcut(self.df[col], q=n_bins,
                                           labels=False, duplicates='drop')
        return self.df[f'{col}_binned']

    def encode_categorical(self, col: str, method: str = 'onehot') -> pd.DataFrame:
        """Encode categorical variables"""
        if method == 'label':
            le = LabelEncoder()
            self.df[f'{col}_encoded'] = le.fit_transform(self.df[col])
            self.transformers[col] = le

        elif method == 'onehot':
            dummies = pd.get_dummies(self.df[col], prefix=col, drop_first=True)
            self.df = pd.concat([self.df, dummies], axis=1)

        return self.df

    def scale_features(self, columns: List[str],
                      method: str = 'standard') -> pd.DataFrame:
        """Scale numeric features"""
        if method == 'standard':
            scaler = StandardScaler()
        elif method == 'minmax':
            from sklearn.preprocessing import MinMaxScaler
            scaler = MinMaxScaler()

        self.df[columns] = scaler.fit_transform(self.df[columns])
        self.transformers['scaler'] = scaler

        return self.df

    def select_features(self, X: pd.DataFrame, y: pd.Series,
                       k: int = 10,
                       method: str = 'f_classif') -> List[str]:
        """Select top k features"""
        if method == 'f_classif':
            scorer = f_classif
        elif method == 'mutual_info':
            scorer = mutual_info_classif

        selector = SelectKBest(scorer, k=k)
        selector.fit(X, y)

        selected_features = X.columns[selector.get_support()].tolist()
        return selected_features
```

## Time Series Analysis

```python
from statsmodels.tsa.seasonal import seasonal_decompose
from statsmodels.tsa.stattools import adfuller
from statsmodels.tsa.arima.model import ARIMA

class TimeSeriesAnalyzer:
    """Analyze time series data"""

    def __init__(self, data: pd.Series, freq: str = 'D'):
        self.data = data
        self.freq = freq

    def decompose(self, model: str = 'additive'):
        """Decompose time series"""
        result = seasonal_decompose(self.data, model=model, period=30)

        return {
            'trend': result.trend,
            'seasonal': result.seasonal,
            'residual': result.resid
        }

    def test_stationarity(self) -> dict:
        """Test for stationarity using Augmented Dickey-Fuller"""
        result = adfuller(self.data.dropna())

        return {
            'adf_statistic': result[0],
            'p_value': result[1],
            'critical_values': result[4],
            'is_stationary': result[1] < 0.05
        }

    def make_stationary(self, method: str = 'diff') -> pd.Series:
        """Make series stationary"""
        if method == 'diff':
            return self.data.diff().dropna()
        elif method == 'log':
            return np.log(self.data)
        elif method == 'log_diff':
            return np.log(self.data).diff().dropna()

    def fit_arima(self, order: tuple = (1, 1, 1)):
        """Fit ARIMA model"""
        model = ARIMA(self.data, order=order)
        fitted_model = model.fit()

        return {
            'model': fitted_model,
            'aic': fitted_model.aic,
            'bic': fitted_model.bic,
            'summary': fitted_model.summary()
        }

    def forecast(self, model, steps: int = 30) -> pd.Series:
        """Generate forecast"""
        return model.forecast(steps=steps)
```

## A/B Testing

```python
from scipy import stats

class ABTest:
    """Conduct A/B tests"""

    def __init__(self, control: np.ndarray, treatment: np.ndarray):
        self.control = control
        self.treatment = treatment

    def ttest(self) -> dict:
        """Two-sample t-test"""
        statistic, p_value = stats.ttest_ind(self.control, self.treatment)

        # Calculate confidence interval for difference
        diff_mean = self.treatment.mean() - self.control.mean()
        se_diff = np.sqrt(self.control.var()/len(self.control) +
                         self.treatment.var()/len(self.treatment))
        ci_lower = diff_mean - 1.96 * se_diff
        ci_upper = diff_mean + 1.96 * se_diff

        return {
            't_statistic': statistic,
            'p_value': p_value,
            'mean_control': self.control.mean(),
            'mean_treatment': self.treatment.mean(),
            'difference': diff_mean,
            'ci_95': (ci_lower, ci_upper),
            'significant': p_value < 0.05
        }

    def proportion_test(self, conversions_control: int,
                       conversions_treatment: int) -> dict:
        """Test difference in proportions"""
        n_control = len(self.control)
        n_treatment = len(self.treatment)

        p_control = conversions_control / n_control
        p_treatment = conversions_treatment / n_treatment

        p_pooled = (conversions_control + conversions_treatment) / (n_control + n_treatment)

        se = np.sqrt(p_pooled * (1 - p_pooled) * (1/n_control + 1/n_treatment))
        z = (p_treatment - p_control) / se
        p_value = 2 * (1 - stats.norm.cdf(abs(z)))

        return {
            'conversion_rate_control': p_control,
            'conversion_rate_treatment': p_treatment,
            'lift': (p_treatment - p_control) / p_control * 100,
            'z_statistic': z,
            'p_value': p_value,
            'significant': p_value < 0.05
        }
```

## Best Practices

### Data Analysis

- Always explore data before modeling
- Check data quality and missing values
- Understand variable distributions
- Look for correlations and relationships
- Document data cleaning steps
- Validate assumptions

### Feature Engineering

- Create domain-specific features
- Test feature importance
- Avoid data leakage
- Use cross-validation for validation
- Document feature transformations
- Keep features interpretable

### Visualization

- Choose appropriate plot types
- Use clear labels and titles
- Consider color accessibility
- Avoid chartjunk
- Tell a story with data
- Make visualizations reproducible

## Anti-Patterns

❌ Not exploring data before modeling
❌ Ignoring data quality issues
❌ Data leakage in feature engineering
❌ Over-engineering features
❌ Misleading visualizations
❌ Not documenting analysis steps
❌ Ignoring business context

## Resources

- Pandas: https://pandas.pydata.org/
- NumPy: https://numpy.org/
- Scikit-learn: https://scikit-learn.org/
- Seaborn: https://seaborn.pydata.org/
- Plotly: https://plotly.com/python/