Temporal Financial Distress Prediction Dataset

Overview

This repository contains sample data from a temporal dataset designed for the early prediction of financial distress among Chinese listed companies. The dataset captures dynamic financial indicators over time, enabling the modeling of temporal relationships and trends inherent in financial metrics. The complete dataset will be made publicly available at a later stage.

Data Files

step_samples.csv

• Description: Contains 50 rows of single time-step sample data, selected at intervals of 3, 6, 9, 12, and 15 months prior to the final prediction point. Each interval includes 10 samples, with an equal split between financial distress samples (label 1) and non-financial distress samples (label 0).

sequence_samples.pkl

• Description: Contains sequences composed of 12 or more time steps of financial ratio data. A balanced set of sequences is included, with equal numbers of labels 0 (non-distress) and 1 (financial distress).

Data Description

The dataset includes 35 financial ratios calculated from financial statements. These ratios cover various aspects of a company's financial health and are essential features for machine learning models aimed at predicting financial distress.

The financial ratios (columns starting with F0...) are defined as follows:

Column	Description	Formula
F030101A	Current Assets Ratio	Total Current Assets ÷ Total Assets
F030201A	Cash Assets Ratio	Ending Balance of Cash and Cash Equivalents ÷ Total Assets
F030301A	Receivables Assets Ratio	(Net Notes Receivable + Net Accounts Receivable) ÷ Total Assets
F030401A	Working Capital to Current Assets Ratio	(Total Current Assets - Total Current Liabilities) ÷ Total Current Assets
F030501A	Working Capital Ratio	(Current Assets - Current Liabilities) ÷ Total Assets
F030601A	Working Capital to Net Assets Ratio	Working Capital ÷ Net Assets
F030701A	Non-Current Assets Ratio	Non-Current Assets ÷ Total Assets
F030801A	Fixed Assets Ratio	Net Fixed Assets ÷ Total Assets
F030901A	Intangible Assets Ratio	Net Intangible Assets ÷ Total Assets
F031001A	Tangible Assets Ratio	Total Tangible Assets ÷ Total Assets
F031101A	Shareholders' Equity Ratio	Total Shareholders' Equity ÷ Total Assets
F031201A	Retained Earnings to Assets Ratio	(Surplus Reserve + Undistributed Profits) ÷ Total Assets
F031301A	Long-term Assets Fit Ratio	(Total Equity + Non-Current Liabilities) ÷ (Net Fixed Assets + Net Available-for-Sale Financial Assets + Net Held-to-Maturity Investments + Net Long-term Equity Investments)
F031401A	Shareholders' Equity to Fixed Assets Ratio	Shareholders' Equity ÷ Net Fixed Assets
F031501A	Current Liabilities Ratio	Total Current Liabilities ÷ Total Liabilities
F031601A	Operating Liabilities Ratio	(Total Current Liabilities - Short-term Borrowings - Non-current Liabilities due within One Year - Trading Financial Liabilities - Derivative Financial Liabilities) ÷ Total Liabilities
F031701A	Financial Liabilities Ratio	(Non-current Liabilities + Short-term Borrowings + Non-current Liabilities due within One Year + Trading Financial Liabilities + Derivative Financial Liabilities) ÷ Total Liabilities
F031801A	Non-Current Liabilities Ratio	Non-Current Liabilities ÷ Total Liabilities
F031901A	Parent Company Shareholders' Equity Ratio	Total Equity Attributable to Parent Company ÷ Total Shareholders' Equity
F032001A	Minority Shareholders' Equity Ratio	Minority Shareholders' Equity ÷ Total Shareholders' Equity
F032101B	Main Business Profit Ratio	(Operating Income - Operating Costs) ÷ Total Profit
F032201B	Financial Activities Profit Ratio	(Investment Income + Fair Value Change Income + Exchange Income) ÷ Total Profit
F032301B	Operating Profit Ratio	Operating Profit ÷ Total Profit
F032401B	Non-operating Income Ratio	(Non-operating Income - Non-operating Expenses) ÷ Total Profit
F032501B	Turnover Tax Rate	Business Taxes and Surcharges ÷ Total Operating Income
F032601B	Comprehensive Tax Rate A	(Business Taxes and Surcharges + Income Tax Expense) ÷ Total Operating Income
F032701B	Comprehensive Tax Rate B	(Business Taxes and Surcharges + Income Tax Expense) ÷ Total Profit
F032801B	Income Tax Rate	Income Tax Expense ÷ Total Profit
F032901B	Net Profit Attributable to Parent Company Ratio	Net Profit Attributable to Parent Company ÷ Net Profit
F033001B	Net Profit Attributable to Minority Shareholders Ratio	Net Profit Attributable to Minority Shareholders ÷ Net Profit
F033101B	Net Profit to Comprehensive Income Ratio	Net Profit ÷ Total Comprehensive Income
F033201B	Other Comprehensive Income Ratio	Other Comprehensive Income ÷ Total Comprehensive Income
F033301B	Comprehensive Income Attributable to Parent Company Ratio	Total Comprehensive Income Attributable to Parent Company ÷ Total Comprehensive Income
F033401B	Comprehensive Income Attributable to Minority Shareholders Ratio	Total Comprehensive Income Attributable to Minority Shareholders ÷ Total Comprehensive Income
F033501A	Parent Company Equity to Invested Capital Ratio	(Total Equity Attributable to Parent Company) ÷ (Total Assets - Total Current Liabilities + Notes Payable + Short-term Borrowings + Non-current Liabilities due within One Year)

Usage

Code to Read and Display step_samples.csv

import pandas as pd

# Read the step_samples.csv file
step_samples = pd.read_csv('step_samples.csv')

# Display the first few rows of the data
print("Step Samples:")
print(step_samples.head())

# If you want to display all rows, use:
# print(step_samples)

Code to Read and Display sequence_samples.pkl

import pickle
import numpy as np

# Load the sequence_samples.pkl file
with open('sequence_samples.pkl', 'rb') as f:
    sequence_samples = pickle.load(f)

# Display the first few samples
print("Sequence Samples:")
for i, (sequence_data, label) in enumerate(sequence_samples[:5]):
    print(f"Sample {i + 1}:")
    print(f"Sequence data shape: {sequence_data.shape}")
    print("Sequence data:")
    print(sequence_data)
    print(f"Label: {label}")
    print("-" * 50)

Notes

• The data provided is for research and educational purposes.
• Labels:
  • 0: Non-financial distress.
  • 1: Financial distress.
• Ensure you have the necessary packages (pandas, numpy, pickle) installed to run the code snippets.

Citation

If you use this dataset in your research, please cite accordingly.

License

This project is licensed under the MIT License.