File size: 14,191 Bytes
f19eb84 | 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 | """
Synthetic advertising dataset.
Generates realistic user profiles, ad inventory, and impression logs
using a latent-factor model to simulate interest-based click / conversion
probabilities that mirror real-world RTB (Real-Time Bidding) behaviour.
Key domain concepts implemented
--------------------------------
* Users – demographic profile + interest vector
* Ads – advertiser, category, format, bid price, targeting
* Impressions – (user, ad, timestamp, clicked, converted, dwell_time, revenue)
* Fatigue – click-probability decays with repeated exposure to same ad
* Reward – composite: click + conversion bonus − fatigue penalty + revenue signal
"""
import os
import numpy as np
import pandas as pd
# ---------------------------------------------------------------------------
# Constants
# ---------------------------------------------------------------------------
CATEGORIES = ['tech', 'fashion', 'sports', 'travel', 'food',
'finance', 'health', 'gaming', 'music', 'automotive']
ADVERTISERS = ['TechCorp', 'FashionHub', 'SportsPro', 'TravelEase',
'FoodieDeals', 'FinancePlus', 'HealthFirst', 'GamingZone',
'MusicStream', 'AutoDrive']
AD_FORMATS = ['banner', 'video', 'native', 'carousel']
AGE_GROUPS = ['18-24', '25-34', '35-44', '45-54', '55+']
GENDERS = ['M', 'F', 'Other']
DEVICES = ['mobile', 'desktop', 'tablet']
# reward weights
R_CLICK = 1.0
R_CONVERT = 3.0
R_FATIGUE = 0.2 # subtracted per repeated impression beyond 2
R_REVENUE = 0.001 # per dollar of bid price
# ---------------------------------------------------------------------------
class AdDataset:
def __init__(self, data_dir: str = 'data', n_users: int = 300, n_ads: int = 100):
self.data_dir = data_dir
self.n_users = n_users
self.n_ads = n_ads
os.makedirs(data_dir, exist_ok=True)
users_path = os.path.join(data_dir, 'users.csv')
ads_path = os.path.join(data_dir, 'ads.csv')
imp_path = os.path.join(data_dir, 'impressions.csv')
if all(os.path.exists(p) for p in (users_path, ads_path, imp_path)):
self.users_df = pd.read_csv(users_path)
self.ads_df = pd.read_csv(ads_path)
self.impressions_df = pd.read_csv(imp_path)
else:
self._generate()
self._preprocess()
# ------------------------------------------------------------------
# Data generation
# ------------------------------------------------------------------
def _generate(self):
np.random.seed(42)
# ── Users ──────────────────────────────────────────────────────
users = []
for uid in range(self.n_users):
n_interests = np.random.randint(1, 4)
interests = np.random.choice(CATEGORIES, size=n_interests, replace=False)
users.append({
'user_id': uid,
'age_group': np.random.choice(AGE_GROUPS),
'gender': np.random.choice(GENDERS),
'interests': '|'.join(interests),
'device': np.random.choice(DEVICES),
})
self.users_df = pd.DataFrame(users)
# ── Ads ────────────────────────────────────────────────────────
ads = []
for ad_id in range(self.n_ads):
cat = CATEGORIES[ad_id % len(CATEGORIES)]
adv = ADVERTISERS[ad_id % len(ADVERTISERS)]
fmt = np.random.choice(AD_FORMATS)
bid = round(np.random.uniform(0.5, 8.0), 2)
budget = round(np.random.uniform(50, 500), 0)
n_tgt = np.random.randint(2, 5)
target_ages = '|'.join(np.random.choice(AGE_GROUPS, size=n_tgt, replace=False))
headlines = {
'tech': 'Upgrade your tech today!', 'fashion': 'Style up this season!',
'sports': 'Gear up for greatness!', 'travel': 'Explore the world now!',
'food': 'Delicious deals await!', 'finance': 'Grow your wealth today!',
'health': 'Live healthier, longer!', 'gaming': 'Level up your game!',
'music': 'Discover new sounds!', 'automotive': 'Drive your dream car!',
}
ads.append({
'ad_id': ad_id,
'advertiser': adv,
'category': cat,
'format': fmt,
'bid_price': bid,
'daily_budget': budget,
'target_ages': target_ages,
'headline': headlines[cat],
'ctr_base': round(np.random.uniform(0.02, 0.06), 4),
'cvr_base': round(np.random.uniform(0.05, 0.15), 4),
})
self.ads_df = pd.DataFrame(ads)
# ── Impressions ────────────────────────────────────────────────
impressions = []
imp_id = 0
for uid in range(self.n_users):
user = self.users_df.iloc[uid]
u_ints = set(user['interests'].split('|'))
n_imps = np.random.randint(30, 100)
ad_ids = np.random.choice(self.n_ads, size=n_imps, replace=True)
freq_count = {} # ad → exposure count for fatigue
for t, ad_id in enumerate(ad_ids):
ad = self.ads_df.iloc[ad_id]
freq = freq_count.get(ad_id, 0)
freq_count[ad_id] = freq + 1
# click probability: base + interest boost + fatigue decay
match = ad['category'] in u_ints
ctr = ad['ctr_base'] * (3.0 if match else 1.0) * (0.8 ** max(0, freq - 1))
ctr = min(ctr, 0.35)
clicked = int(np.random.rand() < ctr)
cvr = ad['cvr_base'] * (1.5 if match else 1.0) if clicked else 0
converted = int(np.random.rand() < cvr)
dwell = np.random.exponential(15.0) if clicked else np.random.exponential(2.0)
revenue = ad['bid_price'] if clicked else 0.0
# reward
fatigue_pen = max(0, freq - 1)
reward = (
R_CLICK * clicked
+ R_CONVERT * converted
- R_FATIGUE * fatigue_pen
+ R_REVENUE * revenue
)
hour = np.random.randint(0, 24)
dow = np.random.randint(0, 7)
impressions.append({
'impression_id': imp_id,
'user_id': uid,
'ad_id': int(ad_id),
'timestamp': 1_000_000 + uid * 1000 + t,
'clicked': clicked,
'converted': converted,
'dwell_time': round(dwell, 2),
'revenue': round(revenue, 4),
'reward': round(reward, 4),
'freq_count': freq,
'hour_of_day': hour,
'day_of_week': dow,
})
imp_id += 1
self.impressions_df = pd.DataFrame(impressions)
# Save
self.users_df.to_csv(os.path.join(self.data_dir, 'users.csv'), index=False)
self.ads_df.to_csv(os.path.join(self.data_dir, 'ads.csv'), index=False)
self.impressions_df.to_csv(os.path.join(self.data_dir, 'impressions.csv'), index=False)
# ------------------------------------------------------------------
# Preprocessing
# ------------------------------------------------------------------
def _preprocess(self):
df = self.impressions_df.sort_values('timestamp')
# Build per-user interaction sequences
self.user_sequences: dict = {}
for uid, grp in df.groupby('user_id'):
self.user_sequences[int(uid)] = list(zip(
grp['ad_id'].values.tolist(),
grp['reward'].values.tolist(),
grp['clicked'].values.tolist(),
grp['converted'].values.tolist(),
))
# Train / test split (80 / 20 per user)
train_rows, test_rows = [], []
for uid, seq in self.user_sequences.items():
split = max(1, int(len(seq) * 0.8))
for i, (ad_id, reward, clicked, converted) in enumerate(seq):
row = {'user_id': uid, 'ad_id': ad_id, 'reward': reward,
'clicked': clicked, 'converted': converted}
(train_rows if i < split else test_rows).append(row)
self.train_df = pd.DataFrame(train_rows)
self.test_df = pd.DataFrame(test_rows)
self.n_users_actual = int(df['user_id'].nunique())
self.n_ads_actual = int(df['ad_id'].nunique())
# Pre-compute aggregate analytics
self._compute_analytics()
def _compute_analytics(self):
df = self.impressions_df
total_imp = len(df)
total_clk = df['clicked'].sum()
total_conv = df['converted'].sum()
total_rev = df['revenue'].sum()
self.analytics = {
'total_impressions': int(total_imp),
'total_clicks': int(total_clk),
'total_conversions': int(total_conv),
'total_revenue': round(float(total_rev), 2),
'ctr': round(float(total_clk / max(1, total_imp)), 4),
'cvr': round(float(total_conv / max(1, total_clk)), 4),
'ecpm': round(float(total_rev / max(1, total_imp) * 1000), 4),
}
# By category
merged = df.merge(self.ads_df[['ad_id', 'category', 'advertiser', 'bid_price']],
on='ad_id', how='left')
cat_stats = []
for cat, g in merged.groupby('category'):
cat_stats.append({
'category': cat,
'impressions': int(len(g)),
'clicks': int(g['clicked'].sum()),
'conversions': int(g['converted'].sum()),
'revenue': round(float(g['revenue'].sum()), 2),
'ctr': round(float(g['clicked'].mean()), 4),
})
self.analytics['by_category'] = cat_stats
# By advertiser
adv_stats = []
for adv, g in merged.groupby('advertiser'):
spend = g['revenue'].sum()
conv = g['converted'].sum()
adv_stats.append({
'advertiser': adv,
'impressions': int(len(g)),
'clicks': int(g['clicked'].sum()),
'conversions': int(conv),
'spend': round(float(spend), 2),
'ctr': round(float(g['clicked'].mean()), 4),
'roas': round(float(conv * 50 / max(0.01, spend)), 2),
})
self.analytics['by_advertiser'] = adv_stats
# ------------------------------------------------------------------
# Helpers
# ------------------------------------------------------------------
def get_user_history(self, user_id: int, max_len: int = 20) -> list:
"""Returns [(ad_id, reward, clicked, converted), ...]."""
return self.user_sequences.get(user_id, [])[-max_len:]
def get_user_features(self, user_id: int) -> np.ndarray:
"""Return a 21-dim binary/float feature vector for the user."""
row = self.users_df[self.users_df['user_id'] == user_id]
if len(row) == 0:
return np.zeros(21, dtype=np.float32)
r = row.iloc[0]
age_oh = np.zeros(5); age_oh[AGE_GROUPS.index(r['age_group'])] = 1
gen_oh = np.zeros(3); gen_oh[GENDERS.index(r['gender'])] = 1
int_oh = np.zeros(10)
for interest in str(r['interests']).split('|'):
if interest in CATEGORIES:
int_oh[CATEGORIES.index(interest)] = 1
dev_oh = np.zeros(3); dev_oh[DEVICES.index(r['device'])] = 1
return np.concatenate([age_oh, gen_oh, int_oh, dev_oh]).astype(np.float32)
def get_context_features(self, hour: int = None, dow: int = None) -> np.ndarray:
"""4-dim sinusoidal encoding of hour and day-of-week."""
if hour is None:
import datetime
now = datetime.datetime.now()
hour = now.hour
dow = now.weekday()
return np.array([
np.sin(2 * np.pi * hour / 24),
np.cos(2 * np.pi * hour / 24),
np.sin(2 * np.pi * dow / 7),
np.cos(2 * np.pi * dow / 7),
], dtype=np.float32)
def get_ad_info(self, ad_id: int) -> dict:
row = self.ads_df[self.ads_df['ad_id'] == ad_id]
if len(row) == 0:
return {'ad_id': ad_id, 'advertiser': 'Unknown', 'category': 'unknown',
'format': 'banner', 'bid_price': 1.0, 'headline': '—'}
r = row.iloc[0]
return {
'ad_id': int(ad_id),
'advertiser': str(r['advertiser']),
'category': str(r['category']),
'format': str(r['format']),
'bid_price': float(r['bid_price']),
'headline': str(r['headline']),
'ctr_base': float(r['ctr_base']),
'cvr_base': float(r['cvr_base']),
}
def get_user_profile(self, user_id: int) -> dict:
row = self.users_df[self.users_df['user_id'] == user_id]
if len(row) == 0:
return {}
r = row.iloc[0]
return {
'user_id': int(user_id),
'age_group': str(r['age_group']),
'gender': str(r['gender']),
'interests': str(r['interests']).split('|'),
'device': str(r['device']),
}
|