File size: 20,752 Bytes
7e960ca |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 |
"""
EDA Analysis Module
This module handles all dataset processing and analysis, providing structured information
about the dataset that can be used for visualization and LLM prompting.
"""
import pandas as pd
import numpy as np
from typing import Dict, List, Tuple, Any, Optional
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.preprocessing import StandardScaler
from io import BytesIO
import base64
class DatasetAnalyzer:
"""Class for analyzing datasets and extracting key information"""
def __init__(self, df: pd.DataFrame = None):
"""Initialize with an optional dataframe"""
self.df = df
self.analysis_results = {}
def load_dataframe(self, df: pd.DataFrame) -> None:
"""Load a dataframe for analysis"""
self.df = df
# Reset analysis results when loading a new dataframe
self.analysis_results = {}
def analyze_dataset(self) -> Dict[str, Any]:
"""
Perform comprehensive analysis on the dataset
Returns:
Dict: Dictionary containing all analysis results
"""
if self.df is None:
raise ValueError("No dataframe loaded. Please load a dataframe first.")
# Basic information
self.analysis_results["shape"] = self.df.shape
self.analysis_results["columns"] = list(self.df.columns)
self.analysis_results["dtypes"] = {col: str(self.df[col].dtype) for col in self.df.columns}
# Missing values
self.analysis_results["missing_values"] = self._analyze_missing_values()
# Basic statistics
self.analysis_results["basic_stats"] = self._generate_basic_stats()
# Correlations (for numerical columns)
self.analysis_results["correlations"] = self._analyze_correlations()
# Sample data
self.analysis_results["sample_data"] = self.df.head().to_string()
# Additional analyses
self.analysis_results["categorical_columns"] = self._identify_categorical_columns()
self.analysis_results["numerical_columns"] = self._identify_numerical_columns()
self.analysis_results["unique_values"] = self._count_unique_values()
return self.analysis_results
def _analyze_missing_values(self) -> Dict[str, Tuple[int, float]]:
"""
Analyze missing values in the dataset
Returns:
Dict: Column names as keys, tuples of (count, percentage) as values
"""
missing_values = {}
for col in self.df.columns:
count = self.df[col].isna().sum()
percentage = round((count / len(self.df)) * 100, 2)
missing_values[col] = (count, percentage)
return missing_values
def _generate_basic_stats(self) -> str:
"""
Generate basic statistics for the dataset
Returns:
str: String representation of basic statistics
"""
# For numerical columns
num_stats = self.df.describe().to_string()
# For categorical columns
cat_columns = self._identify_categorical_columns()
cat_stats = ""
if cat_columns:
cat_stats = "\n\nCategorical columns statistics:\n"
for col in cat_columns:
value_counts = self.df[col].value_counts().head(10)
cat_stats += f"\n{col} - Top values:\n{value_counts.to_string()}\n"
return num_stats + cat_stats
def _analyze_correlations(self) -> str:
"""
Analyze correlations between numerical features
Returns:
str: String representation of top correlations
"""
num_columns = self._identify_numerical_columns()
if not num_columns or len(num_columns) < 2:
return "Not enough numerical columns for correlation analysis."
corr_matrix = self.df[num_columns].corr()
# Get top correlations (excluding self-correlations)
corr_pairs = []
for i in range(len(num_columns)):
for j in range(i+1, len(num_columns)):
col1, col2 = num_columns[i], num_columns[j]
corr_value = corr_matrix.loc[col1, col2]
if not np.isnan(corr_value):
corr_pairs.append((col1, col2, corr_value))
# Sort by absolute correlation value
corr_pairs.sort(key=lambda x: abs(x[2]), reverse=True)
# Format results
result = "Top correlations:\n"
for col1, col2, corr in corr_pairs[:10]: # Top 10 correlations
result += f"{col1} -- {col2}: {corr:.4f}\n"
return result
def _identify_categorical_columns(self) -> List[str]:
"""
Identify categorical columns in the dataset
Returns:
List[str]: List of categorical column names
"""
cat_columns = []
for col in self.df.columns:
# Consider object, category, and boolean types as categorical
if self.df[col].dtype == 'object' or self.df[col].dtype == 'category' or self.df[col].dtype == 'bool':
cat_columns.append(col)
# Also consider int/float columns with few unique values as categorical
elif (self.df[col].dtype == 'int64' or self.df[col].dtype == 'float64') and \
self.df[col].nunique() < 10 and self.df[col].nunique() / len(self.df) < 0.05:
cat_columns.append(col)
return cat_columns
def _identify_numerical_columns(self) -> List[str]:
"""
Identify numerical columns in the dataset
Returns:
List[str]: List of numerical column names
"""
num_columns = []
cat_columns = self._identify_categorical_columns()
for col in self.df.columns:
if col not in cat_columns and pd.api.types.is_numeric_dtype(self.df[col].dtype):
num_columns.append(col)
return num_columns
def _count_unique_values(self) -> Dict[str, int]:
"""
Count unique values for each column
Returns:
Dict: Column names as keys, unique count as values
"""
return {col: self.df[col].nunique() for col in self.df.columns}
def generate_eda_visualizations(self) -> Dict[str, str]:
"""
Generate common EDA visualizations
Returns:
Dict: Dictionary of visualization titles and their base64-encoded images
"""
if self.df is None:
raise ValueError("No dataframe loaded. Please load a dataframe first.")
visualizations = {}
# 1. Missing values heatmap
visualizations["missing_values_heatmap"] = self._plot_missing_values()
# 2. Distribution plots for numerical features
num_columns = self._identify_numerical_columns()
for i, col in enumerate(num_columns[:5]): # Limit to first 5 numerical columns
visualizations[f"distribution_{col}"] = self._plot_distribution(col)
# 3. Correlation heatmap
visualizations["correlation_heatmap"] = self._plot_correlation_heatmap()
# 4. Categorical feature distributions
cat_columns = self._identify_categorical_columns()
for i, col in enumerate(cat_columns[:5]): # Limit to first 5 categorical columns
visualizations[f"categorical_{col}"] = self._plot_categorical_distribution(col)
# 5. Scatter plot of 2 most correlated features
if len(num_columns) >= 2:
visualizations["scatter_plot"] = self._plot_scatter_correlation()
return visualizations
def _plot_missing_values(self) -> str:
"""Generate missing values heatmap"""
plt.figure(figsize=(10, 6))
sns.heatmap(self.df.isnull(), cmap='viridis', yticklabels=False, cbar=True, cbar_kws={'label': 'Missing Data'})
plt.tight_layout()
plt.title('Missing Values Heatmap')
# Convert plot to base64 string
return self._fig_to_base64(plt.gcf())
def _plot_distribution(self, column: str) -> str:
"""Generate distribution plot for a numerical column"""
plt.figure(figsize=(10, 6))
# Histogram with KDE
sns.histplot(data=self.df, x=column, kde=True)
plt.title(f'Distribution of {column}')
plt.xlabel(column)
plt.ylabel('Frequency')
plt.tight_layout()
# Convert plot to base64 string
return self._fig_to_base64(plt.gcf())
def _plot_correlation_heatmap(self) -> str:
"""Generate correlation heatmap"""
num_columns = self._identify_numerical_columns()
if not num_columns or len(num_columns) < 2:
return ""
plt.figure(figsize=(12, 10))
corr_matrix = self.df[num_columns].corr()
mask = np.triu(np.ones_like(corr_matrix, dtype=bool))
# Custom diverging palette
cmap = sns.diverging_palette(230, 20, as_cmap=True)
# Draw heatmap
sns.heatmap(corr_matrix, mask=mask, cmap=cmap, vmax=1, vmin=-1, center=0,
square=True, linewidths=.5, annot=True, fmt=".2f")
plt.title('Correlation Heatmap')
plt.tight_layout()
# Convert plot to base64 string
return self._fig_to_base64(plt.gcf())
def _plot_categorical_distribution(self, column: str) -> str:
"""Generate bar plot for categorical column"""
plt.figure(figsize=(10, 6))
# Get value counts and limit to top 10 categories if there are too many
value_counts = self.df[column].value_counts()
if len(value_counts) > 10:
# Keep top 9 categories and group the rest as 'Other'
top_categories = value_counts.nlargest(9).index
data = self.df.copy()
data[column] = data[column].apply(lambda x: x if x in top_categories else 'Other')
sns.countplot(y=column, data=data, order=data[column].value_counts().index)
else:
sns.countplot(y=column, data=self.df, order=value_counts.index)
plt.title(f'Distribution of {column}')
plt.xlabel('Count')
plt.ylabel(column)
plt.tight_layout()
# Convert plot to base64 string
return self._fig_to_base64(plt.gcf())
def _plot_scatter_correlation(self) -> str:
"""Generate scatter plot of two most correlated features"""
num_columns = self._identify_numerical_columns()
if not num_columns or len(num_columns) < 2:
return ""
# Find the two most correlated features
corr_matrix = self.df[num_columns].corr().abs()
# Get upper triangle mask
mask = np.triu(np.ones_like(corr_matrix, dtype=bool))
corr_matrix = corr_matrix.mask(mask)
# Find the max correlation
max_corr = corr_matrix.max().max()
max_corr_idx = corr_matrix.stack().idxmax()
if pd.isna(max_corr):
return ""
# Get the column names
col1, col2 = max_corr_idx
# Create scatter plot
plt.figure(figsize=(10, 6))
# Add regression line
sns.regplot(x=col1, y=col2, data=self.df, scatter_kws={'alpha': 0.5})
plt.title(f'Scatter plot of {col1} vs {col2} (correlation: {corr_matrix.loc[col1, col2]:.2f})')
plt.tight_layout()
# Convert plot to base64 string
return self._fig_to_base64(plt.gcf())
def _fig_to_base64(self, fig) -> str:
"""Convert matplotlib figure to base64 string"""
buf = BytesIO()
fig.savefig(buf, format='png', bbox_inches='tight')
buf.seek(0)
img_str = base64.b64encode(buf.read()).decode('utf-8')
plt.close(fig)
return img_str
def suggest_data_preprocessing(self) -> Dict[str, List[str]]:
"""
Suggest preprocessing steps based on dataset analysis
Returns:
Dict: Dictionary of preprocessing suggestions for each column type
"""
if not self.analysis_results:
self.analyze_dataset()
suggestions = {
"numerical": [],
"categorical": [],
"missing_values": [],
"outliers": [],
"general": []
}
# Missing values suggestions
missing_cols = [col for col, (count, _) in self.analysis_results["missing_values"].items() if count > 0]
if missing_cols:
suggestions["missing_values"].append(f"Found {len(missing_cols)} columns with missing values.")
if len(missing_cols) > 5:
suggestions["missing_values"].append(f"Columns with highest missing values: {', '.join(missing_cols[:5])}...")
else:
suggestions["missing_values"].append(f"Columns with missing values: {', '.join(missing_cols)}")
suggestions["missing_values"].append("Consider these strategies for handling missing values:")
suggestions["missing_values"].append("- Imputation (mean/median for numerical, mode for categorical)")
suggestions["missing_values"].append("- Creating missing value indicators as new features")
suggestions["missing_values"].append("- Removing rows or columns with too many missing values")
# Numerical column suggestions
num_cols = self.analysis_results["numerical_columns"]
if num_cols:
suggestions["numerical"].append(f"Found {len(num_cols)} numerical columns.")
suggestions["numerical"].append("Consider these preprocessing steps:")
suggestions["numerical"].append("- Scaling (StandardScaler or MinMaxScaler)")
suggestions["numerical"].append("- Check for skewness and apply log or Box-Cox transformation if needed")
suggestions["numerical"].append("- Create binned versions of continuous variables")
# Check for potential outliers
for col in num_cols:
if col in self.df.columns: # Safety check
q1 = self.df[col].quantile(0.25)
q3 = self.df[col].quantile(0.75)
iqr = q3 - q1
outlier_count = ((self.df[col] < (q1 - 1.5 * iqr)) | (self.df[col] > (q3 + 1.5 * iqr))).sum()
if outlier_count > 0:
percentage = round((outlier_count / len(self.df)) * 100, 2)
if percentage > 5: # If more than 5% are outliers
suggestions["outliers"].append(f"Column '{col}' has {outlier_count} potential outliers ({percentage}%).")
# Categorical column suggestions
cat_cols = self.analysis_results["categorical_columns"]
if cat_cols:
suggestions["categorical"].append(f"Found {len(cat_cols)} categorical columns.")
# Check cardinality (number of unique values)
high_cardinality = []
for col in cat_cols:
unique_count = self.analysis_results["unique_values"].get(col, 0)
if unique_count > 10:
high_cardinality.append((col, unique_count))
if high_cardinality:
suggestions["categorical"].append("High cardinality columns (many unique values):")
for col, count in sorted(high_cardinality, key=lambda x: x[1], reverse=True)[:5]:
suggestions["categorical"].append(f"- {col}: {count} unique values")
suggestions["categorical"].append("For high cardinality columns, consider:")
suggestions["categorical"].append("- Grouping less frequent categories")
suggestions["categorical"].append("- Target encoding or embedding techniques")
suggestions["categorical"].append("General categorical encoding strategies:")
suggestions["categorical"].append("- One-hot encoding for low cardinality columns")
suggestions["categorical"].append("- Label encoding for ordinal variables")
# General suggestions
suggestions["general"].append("General preprocessing recommendations:")
suggestions["general"].append("- Check for duplicate rows and remove if necessary")
suggestions["general"].append("- Normalize text fields (lowercase, remove special characters)")
suggestions["general"].append("- Create feature interactions for highly correlated features")
return suggestions
def generate_feature_engineering_ideas(self) -> List[str]:
"""
Generate feature engineering ideas based on dataset analysis
Returns:
List[str]: List of feature engineering suggestions
"""
if not self.analysis_results:
self.analyze_dataset()
ideas = []
# Get column types
num_cols = self.analysis_results["numerical_columns"]
cat_cols = self.analysis_results["categorical_columns"]
# Aggregation features
if len(num_cols) >= 2:
ideas.append("### Numerical Feature Transformations:")
ideas.append("1. Create polynomial features for continuous variables")
ideas.append("2. Apply mathematical transformations (log, sqrt, square) to handle skewed distributions")
ideas.append("3. Create binned versions of continuous features to capture non-linear relationships")
# Check for date/time related column names
time_related_cols = [col for col in self.df.columns if any(x in col.lower() for x in ['date', 'time', 'year', 'month', 'day'])]
if time_related_cols:
ideas.append("\n### Time-Based Features:")
ideas.append(f"Detected potential date/time columns: {', '.join(time_related_cols)}")
ideas.append("1. Extract components like year, month, day, weekday, quarter")
ideas.append("2. Create cyclical features using sine/cosine transformations for periodic time components")
ideas.append("3. Calculate time since specific events or time differences between dates")
# Categorical interactions
if len(cat_cols) >= 2:
ideas.append("\n### Categorical Feature Engineering:")
ideas.append("1. Create interaction features by combining categorical variables")
ideas.append("2. Use target encoding for high cardinality categorical features")
ideas.append("3. Combine rare categories into an 'Other' category to reduce dimensionality")
# Mixed interactions
if num_cols and cat_cols:
ideas.append("\n### Feature Interactions:")
ideas.append("1. Create group-based statistics (mean, median, min, max) of numerical features grouped by categorical features")
ideas.append("2. Calculate the difference from group means for numerical features")
ideas.append("3. Create ratio or difference features between related numerical columns")
# Dimensionality reduction
if len(num_cols) > 10:
ideas.append("\n### Dimensionality Reduction:")
ideas.append("1. Apply PCA to reduce dimensionality and create principal components")
ideas.append("2. Use feature selection methods (information gain, chi-square, mutual information)")
ideas.append("3. Try UMAP or t-SNE for non-linear dimensionality reduction")
# Text features
text_cols = [col for col in self.df.columns if self.df[col].dtype == 'object' and
self.df[col].apply(lambda x: isinstance(x, str) and len(x.split()) > 3).mean() > 0.5]
if text_cols:
ideas.append("\n### Text Feature Engineering:")
ideas.append(f"Detected potential text columns: {', '.join(text_cols)}")
ideas.append("1. Create bag-of-words or TF-IDF representations")
ideas.append("2. Extract text length, word count, and other statistical features")
ideas.append("3. Consider pretrained word embeddings or sentence transformers")
return ideas
|