File size: 15,194 Bytes
791e439 | 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 | """
Fraud Detection Engine - Hugging Face Spaces Version
This version is configured for deployment on Hugging Face Spaces.
Data files should be placed in the 'data/' directory within the Space.
"""
import gradio as gr
import json
import os
import re
import pandas as pd
from datetime import datetime
from typing import Optional, Dict, List, Any
# ============================================================================
# CONFIGURATION
# ============================================================================
# For HF Spaces, use relative paths
BASE_DATA_PATH = os.environ.get("DATA_PATH", "data")
FINANCIAL_DATA_CSV = os.path.join(BASE_DATA_PATH, "Financial Data.csv")
BENEISH_MSCORE_CSV = os.path.join(BASE_DATA_PATH, "Beneish M-score - Sheet1.csv")
ZSCORE_DATA_CSV = os.path.join(BASE_DATA_PATH, "Z-score data.csv")
COMPANY_TICKERS_JSON = os.path.join(BASE_DATA_PATH, "company_tickers.json")
FILINGS_DIR = os.path.join(BASE_DATA_PATH, "filings")
MSCORE_MANIPULATION_THRESHOLD = -1.78
ZSCORE_SAFE_THRESHOLD = 2.99
ZSCORE_GREY_THRESHOLD = 1.81
# ============================================================================
# SIMPLIFIED DATA LOADER
# ============================================================================
class DataLoader:
"""Simplified data loader for HF Spaces deployment."""
_financial_df: Optional[pd.DataFrame] = None
_mscore_df: Optional[pd.DataFrame] = None
_zscore_df: Optional[pd.DataFrame] = None
_company_tickers: Optional[Dict] = None
@classmethod
def _load_financial_csv(cls) -> pd.DataFrame:
if cls._financial_df is None:
if os.path.exists(FINANCIAL_DATA_CSV):
cls._financial_df = pd.read_csv(FINANCIAL_DATA_CSV)
cls._financial_df['cik_normalized'] = pd.to_numeric(
cls._financial_df['cik'], errors='coerce'
).astype('Int64')
else:
cls._financial_df = pd.DataFrame()
return cls._financial_df
@classmethod
def _load_mscore_csv(cls) -> pd.DataFrame:
if cls._mscore_df is None:
if os.path.exists(BENEISH_MSCORE_CSV):
cls._mscore_df = pd.read_csv(BENEISH_MSCORE_CSV)
if 'CIK Numbers' in cls._mscore_df.columns:
cls._mscore_df['cik_normalized'] = pd.to_numeric(
cls._mscore_df['CIK Numbers'], errors='coerce'
).astype('Int64')
else:
cls._mscore_df = pd.DataFrame()
return cls._mscore_df
@classmethod
def _load_zscore_csv(cls) -> pd.DataFrame:
if cls._zscore_df is None:
if os.path.exists(ZSCORE_DATA_CSV):
cls._zscore_df = pd.read_csv(ZSCORE_DATA_CSV)
cls._zscore_df['cik_normalized'] = pd.to_numeric(
cls._zscore_df['cik'], errors='coerce'
).astype('Int64')
else:
cls._zscore_df = pd.DataFrame()
return cls._zscore_df
@classmethod
def load_company_tickers(cls) -> Dict[int, Dict[str, str]]:
if cls._company_tickers is None:
if os.path.exists(COMPANY_TICKERS_JSON):
with open(COMPANY_TICKERS_JSON, 'r') as f:
raw = json.load(f)
cls._company_tickers = {}
for idx, company in raw.items():
cik = company.get('cik_str')
if cik:
cls._company_tickers[int(cik)] = {
"ticker": company.get('ticker', ''),
"title": company.get('title', '')
}
else:
cls._company_tickers = {}
return cls._company_tickers
@classmethod
def get_company_info(cls, cik: int) -> Dict[str, str]:
tickers = cls.load_company_tickers()
return tickers.get(cik, {"ticker": "Unknown", "title": "Unknown"})
@classmethod
def get_available_companies(cls) -> List[tuple]:
"""Get list of companies with available data."""
tickers = cls.load_company_tickers()
mscore_df = cls._load_mscore_csv()
if mscore_df.empty:
# Return sample if no data
return [("DEMO - Sample Company", 0)]
available_ciks = set(mscore_df['cik_normalized'].dropna().astype(int).tolist())
choices = []
for cik in available_ciks:
info = tickers.get(cik, {})
ticker = info.get('ticker', 'UNK')
name = info.get('title', 'Unknown')
if ticker and ticker != 'Unknown':
choices.append((f"{ticker} - {name[:40]}", cik))
return sorted(choices, key=lambda x: x[0]) if choices else [("DEMO - Sample Company", 0)]
@classmethod
def get_precomputed_mscore(cls, cik: int) -> Optional[float]:
df = cls._load_mscore_csv()
if df.empty:
return None
row = df[df['cik_normalized'] == cik]
if row.empty:
return None
m_score = row.iloc[0].get('m_score')
return float(m_score) if pd.notna(m_score) else None
@classmethod
def get_zscore_inputs(cls, cik: int) -> Optional[Dict[str, float]]:
df = cls._load_zscore_csv()
if df.empty:
return None
row = df[df['cik_normalized'] == cik]
if row.empty:
return None
row = row.iloc[-1]
try:
at = float(row.get('at', 0) or 0)
if at == 0:
return None
act = float(row.get('act', 0) or 0)
lct = float(row.get('lct', 0) or 0)
re_val = float(row.get('re', 0) or 0)
ebit = float(row.get('ebit', 0) or 0)
revt = float(row.get('revt', 0) or 0)
csho = float(row.get('csho', 0) or 0)
prcc_f = float(row.get('prcc_f', 0) or 0)
mve = csho * prcc_f
tl = lct if lct > 0 else at * 0.5
return {
"x1": (act - lct) / at,
"x2": re_val / at,
"x3": ebit / at,
"x4": mve / tl if tl > 0 else 0,
"x5": revt / at
}
except (ValueError, TypeError):
return None
# ============================================================================
# AGENTS
# ============================================================================
class FinancialAgent:
def __init__(self):
self.name = "Financial Analysis Agent"
def calculate_beneish_m_score(self, data: dict) -> dict:
try:
m_score = (
-4.84
+ 0.92 * data.get('dsri', 0)
+ 0.52 * data.get('gmi', 0)
+ 0.71 * data.get('aqi', 0)
+ 0.20 * data.get('sgi', 0)
+ 0.11 * data.get('depi', 0)
- 0.17 * data.get('sgai', 0)
+ 4.67 * data.get('tata', 0)
- 0.32 * data.get('lvgi', 0)
)
risk_flag = m_score > MSCORE_MANIPULATION_THRESHOLD
return {
"m_score": round(m_score, 4),
"risk_flag": risk_flag,
"details": "High probability of manipulation" if risk_flag else "Low probability of manipulation"
}
except Exception as e:
return {"error": str(e)}
def calculate_altman_z_score(self, data: dict) -> dict:
try:
z_score = (
1.2 * data.get('x1', 0)
+ 1.4 * data.get('x2', 0)
+ 3.3 * data.get('x3', 0)
+ 0.6 * data.get('x4', 0)
+ 1.0 * data.get('x5', 0)
)
if z_score > ZSCORE_SAFE_THRESHOLD:
status = "Safe Zone"
elif z_score > ZSCORE_GREY_THRESHOLD:
status = "Grey Zone"
else:
status = "Distress Zone"
return {"z_score": round(z_score, 4), "status": status}
except Exception as e:
return {"error": str(e)}
class RiskAgent:
def __init__(self):
self.name = "Risk Assessment Agent"
def calculate_final_risk(self, financial_results: dict, text_results: dict) -> dict:
risk_score = 0
reasons = []
if financial_results.get("risk_flag"):
risk_score += 40
reasons.append("Beneish M-Score indicates manipulation risk")
z_status = financial_results.get("altman_z", {}).get("status")
if z_status == "Distress Zone":
risk_score += 30
reasons.append("Altman Z-Score indicates financial distress")
elif z_status == "Grey Zone":
risk_score += 15
reasons.append("Altman Z-Score in Grey Zone")
obfuscation = text_results.get("obfuscation_score", 0)
if obfuscation > 0.7:
risk_score += 30
reasons.append(f"High managerial obfuscation (Score: {obfuscation:.2f})")
elif obfuscation > 0.4:
risk_score += 10
if risk_score > 70:
risk_level = "CRITICAL"
elif risk_score > 40:
risk_level = "HIGH"
elif risk_score > 20:
risk_level = "MODERATE"
else:
risk_level = "LOW"
return {
"total_risk_score": risk_score,
"risk_level": risk_level,
"key_factors": reasons
}
# ============================================================================
# ANALYSIS FUNCTION
# ============================================================================
fin_agent = FinancialAgent()
risk_agent = RiskAgent()
def analyze_company(cik_selection):
"""Run fraud detection analysis on selected company."""
if not cik_selection:
return "Please select a company", "", "", ""
cik = int(cik_selection)
# Demo mode
if cik == 0:
return demo_analysis()
company_info = DataLoader.get_company_info(cik)
# Financial Analysis
m_score_val = DataLoader.get_precomputed_mscore(cik)
if m_score_val is not None:
m_score_result = {
"m_score": round(m_score_val, 4),
"risk_flag": m_score_val > MSCORE_MANIPULATION_THRESHOLD,
"details": "High probability of manipulation" if m_score_val > MSCORE_MANIPULATION_THRESHOLD else "Low probability of manipulation"
}
else:
m_score_result = {"m_score": None, "risk_flag": False, "details": "Data not available"}
zscore_inputs = DataLoader.get_zscore_inputs(cik)
if zscore_inputs:
z_score_result = fin_agent.calculate_altman_z_score(zscore_inputs)
else:
z_score_result = {"z_score": None, "status": "Unknown"}
financial_results = {
"beneish_m": m_score_result,
"altman_z": z_score_result,
"risk_flag": m_score_result.get("risk_flag", False) or (z_score_result.get("status") == "Distress Zone")
}
# Simplified text results (no 10-K file analysis in HF version)
text_results = {
"obfuscation_score": 0.3, # Placeholder
"note": "Text analysis requires 10-K filing upload"
}
# Risk Assessment
final_report = risk_agent.calculate_final_risk(financial_results, text_results)
# Format outputs
company_header = f"""## {company_info['ticker']} - {company_info['title']}
**CIK:** {cik}
"""
m_val = m_score_result.get('m_score', 'N/A')
m_flag = "HIGH RISK" if m_score_result.get('risk_flag') else "Normal"
z_val = z_score_result.get('z_score', 'N/A')
z_status = z_score_result.get('status', 'Unknown')
financial_output = f"""### Beneish M-Score
**Score:** {m_val}
**Status:** {m_flag}
**Interpretation:** {m_score_result.get('details', 'N/A')}
> M-Score > -1.78 indicates high probability of earnings manipulation
### Altman Z-Score
**Score:** {z_val}
**Status:** {z_status}
> Safe Zone (>2.99) | Grey Zone (1.81-2.99) | Distress Zone (<1.81)
"""
text_output = """### MD&A Analysis
**Status:** Not available in demo
> Upload 10-K filings for full text analysis
"""
risk_level = final_report['risk_level']
risk_score = final_report['total_risk_score']
risk_output = f"""## FRAUD RISK ASSESSMENT
### Risk Level: {risk_level}
### Total Score: {risk_score}/100
**Key Risk Factors:**
"""
factors = final_report.get('key_factors', [])
if factors:
risk_output += "\n".join([f"- {f}" for f in factors])
else:
risk_output += "- No significant risk factors identified"
return company_header, financial_output, text_output, risk_output
def demo_analysis():
"""Return demo analysis when no real data is available."""
company_header = """## DEMO - Sample Analysis
**Note:** This is a demonstration with sample data.
Upload your data files to analyze real companies.
"""
financial_output = """### Beneish M-Score (Demo)
**Score:** -2.45
**Status:** Normal
**Interpretation:** Low probability of manipulation
### Altman Z-Score (Demo)
**Score:** 3.25
**Status:** Safe Zone
"""
text_output = """### MD&A Analysis (Demo)
**Obfuscation Score:** 0.35
**Flagged Phrases:** 3 found
- "challenging market conditions"
- "strategic realignment"
- "factors beyond our control"
"""
risk_output = """## FRAUD RISK ASSESSMENT (Demo)
### Risk Level: LOW
### Total Score: 15/100
**Key Risk Factors:**
- No significant risk factors in this demo
"""
return company_header, financial_output, text_output, risk_output
# ============================================================================
# GRADIO INTERFACE
# ============================================================================
with gr.Blocks(title="Fraud Detection Engine") as demo:
gr.Markdown("""
# Fraud Detection Engine
Analyzes SEC 10-K filings to identify risks of financial statement manipulation using:
- **Beneish M-Score** - Earnings manipulation detection
- **Altman Z-Score** - Financial distress assessment
- **MD&A Text Analysis** - Managerial obfuscation detection
""")
with gr.Row():
company_dropdown = gr.Dropdown(
choices=DataLoader.get_available_companies(),
label="Select Company",
info="Companies with available financial data"
)
analyze_btn = gr.Button("Analyze", variant="primary")
company_info = gr.Markdown()
with gr.Row():
with gr.Column():
financial_output = gr.Markdown(label="Financial Analysis")
with gr.Column():
text_output = gr.Markdown(label="Text Analysis")
risk_output = gr.Markdown()
analyze_btn.click(
fn=analyze_company,
inputs=[company_dropdown],
outputs=[company_info, financial_output, text_output, risk_output]
)
gr.Markdown("""
---
**Methodology:** Based on Beneish (1999) M-Score and Altman (1968) Z-Score models.
**Data Requirements:** Place CSV files in `data/` directory:
- `Financial Data.csv`
- `Beneish M-score - Sheet1.csv`
- `Z-score data.csv`
- `company_tickers.json`
""")
if __name__ == "__main__":
demo.launch(server_name="0.0.0.0", server_port=7860)
|