File size: 7,793 Bytes
198ccb0 | 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 | """Predictive intervals for sentiment analysis using Beta distribution."""
import math
import logging
from typing import List, Dict, Tuple, Optional
import pandas as pd
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
def calculate_predictive_interval(
positive_count: int,
negative_count: int,
neutral_count: int = 0,
confidence_level: float = 0.95
) -> float:
"""
Calculate lower bound of predictive interval for positive comment ratio.
Uses Beta distribution to model the proportion of positive comments.
This accounts for uncertainty when sample size is small.
Formula:
a = 1 + u (positive comments)
b = 1 + d (negative + neutral comments)
Lower bound = mean - z_score * std_dev
Args:
positive_count: Number of positive comments
negative_count: Number of negative comments
neutral_count: Number of neutral comments (default: 0)
confidence_level: Confidence level (0.95 for 95%, 0.99 for 99%)
Returns:
Lower bound of predictive interval (0.0 to 1.0)
Example:
>>> # 80 positive, 20 negative out of 100 comments
>>> lower_bound = calculate_predictive_interval(80, 20)
>>> print(f"Lower bound: {lower_bound:.3f}")
Lower bound: 0.742
"""
u = positive_count
d = negative_count + neutral_count
# Beta distribution parameters
a = 1 + u
b = 1 + d
# Mean of Beta distribution
mean = a / (a + b)
# Variance of Beta distribution
variance = (a * b) / ((a + b) ** 2 * (a + b + 1))
std_dev = math.sqrt(variance)
# Z-score for confidence level
# 95% confidence: z = 1.65 (one-sided)
# 99% confidence: z = 2.33 (one-sided)
z_scores = {
0.90: 1.28,
0.95: 1.65,
0.99: 2.33
}
z_score = z_scores.get(confidence_level, 1.65)
# Lower bound of predictive interval
lower_bound = mean - z_score * std_dev
# Ensure non-negative and within [0, 1]
lower_bound = max(0.0, min(1.0, lower_bound))
return lower_bound
def rank_by_predictive_interval(
data: List[Dict],
positive_key: str = "positive_count",
negative_key: str = "negative_count",
neutral_key: str = "neutral_count",
confidence_level: float = 0.95
) -> List[Dict]:
"""
Rank items by predictive interval lower bound.
This is useful for ranking news articles or categories by positive
sentiment while accounting for sample size uncertainty.
Args:
data: List of dictionaries with sentiment counts
positive_key: Key for positive count in data dict
negative_key: Key for negative count in data dict
neutral_key: Key for neutral count in data dict
confidence_level: Confidence level for interval
Returns:
List of dictionaries sorted by predictive interval (descending)
Each dict includes 'predictive_interval' field
Example:
>>> data = [
... {"id": 1, "positive_count": 80, "negative_count": 20},
... {"id": 2, "positive_count": 1, "negative_count": 0},
... ]
>>> ranked = rank_by_predictive_interval(data)
>>> ranked[0]["id"] # First item has higher interval
1
"""
results = []
for item in data:
positive = item.get(positive_key, 0)
negative = item.get(negative_key, 0)
neutral = item.get(neutral_key, 0)
interval = calculate_predictive_interval(
positive_count=positive,
negative_count=negative,
neutral_count=neutral,
confidence_level=confidence_level
)
# Create new dict with interval
result = item.copy()
result["predictive_interval"] = interval
result["total_comments"] = positive + negative + neutral
result["positive_ratio"] = positive / (positive + negative + neutral) if (positive + negative + neutral) > 0 else 0.0
results.append(result)
# Sort by predictive interval (descending)
results.sort(key=lambda x: x["predictive_interval"], reverse=True)
return results
def calculate_intervals_for_dataframe(
df: pd.DataFrame,
positive_col: str = "positive_count",
negative_col: str = "negative_count",
neutral_col: str = "neutral_count",
confidence_level: float = 0.95
) -> pd.DataFrame:
"""
Calculate predictive intervals for DataFrame.
Args:
df: DataFrame with sentiment counts
positive_col: Column name for positive counts
negative_col: Column name for negative counts
neutral_col: Column name for neutral counts
confidence_level: Confidence level
Returns:
DataFrame with added 'predictive_interval' column
Example:
>>> df = pd.DataFrame({
... "positive_count": [80, 1],
... "negative_count": [20, 0]
... })
>>> df_with_intervals = calculate_intervals_for_dataframe(df)
>>> "predictive_interval" in df_with_intervals.columns
True
"""
df = df.copy()
df["predictive_interval"] = df.apply(
lambda row: calculate_predictive_interval(
positive_count=row.get(positive_col, 0),
negative_count=row.get(negative_col, 0),
neutral_count=row.get(neutral_col, 0),
confidence_level=confidence_level
),
axis=1
)
return df
def get_top_positive_by_interval(
data: List[Dict],
top_k: int = 10,
min_comments: int = 1,
**kwargs
) -> List[Dict]:
"""
Get top K items ranked by predictive interval.
Args:
data: List of dictionaries with sentiment counts
top_k: Number of top items to return
min_comments: Minimum number of comments required
**kwargs: Additional arguments for rank_by_predictive_interval
Returns:
Top K items sorted by predictive interval
Example:
>>> data = [
... {"id": 1, "positive_count": 80, "negative_count": 20},
... {"id": 2, "positive_count": 1, "negative_count": 0},
... ]
>>> top = get_top_positive_by_interval(data, top_k=1)
>>> len(top)
1
"""
# Filter by minimum comments
filtered = [
item for item in data
if (item.get("positive_count", 0) +
item.get("negative_count", 0) +
item.get("neutral_count", 0)) >= min_comments
]
# Rank by predictive interval
ranked = rank_by_predictive_interval(filtered, **kwargs)
# Return top K
return ranked[:top_k]
def get_top_negative_by_interval(
data: List[Dict],
top_k: int = 10,
min_comments: int = 1,
**kwargs
) -> List[Dict]:
"""
Get top K items ranked by negative sentiment (lowest predictive interval).
Args:
data: List of dictionaries with sentiment counts
top_k: Number of top items to return
min_comments: Minimum number of comments required
**kwargs: Additional arguments for rank_by_predictive_interval
Returns:
Top K items with lowest predictive intervals (most negative)
"""
# Filter by minimum comments
filtered = [
item for item in data
if (item.get("positive_count", 0) +
item.get("negative_count", 0) +
item.get("neutral_count", 0)) >= min_comments
]
# Rank by predictive interval
ranked = rank_by_predictive_interval(filtered, **kwargs)
# Return bottom K (most negative)
return ranked[-top_k:][::-1] # Reverse to get most negative first
|