electricsheepafrica/Africa-Hepatitis-C-prediction-dataset

This dataset contains laboratory values of blood donors and Hepatitis C patients with demographic information. The dataset provides comprehensive medical data for Hepatitis C research and machine learning applications.

Dataset Summary

• Total Samples: 615
• Features: 13 (Patient ID, Category, Age, Sex, 9 laboratory values)
• Task: Multi-class classification of Hepatitis C diagnosis and disease progression
• Format: Parquet files for efficient storage and loading

Features

Patient Information

• Unnamed: 0: Patient ID/Number (integer)
• Category: diagnosis category (string)
  • 0=Blood Donor: Healthy blood donor
  • 0s=suspect Blood Donor: Suspect blood donor
  • 1=Hepatitis: Hepatitis C infection
  • 2=Fibrosis: Liver fibrosis
  • 3=Cirrhosis: Liver cirrhosis
• Age: Age in years (integer)
• Sex: Gender (f=female, m=male)

Laboratory Values

• ALB: Albumin (g/dL) - Liver function marker
• ALP: Alkaline phosphatase (U/L) - Bone and liver enzyme
• ALT: Alanine aminotransferase (U/L) - Liver enzyme
• AST: Aspartate aminotransferase (U/L) - Liver enzyme
• BIL: Bilirubin (mg/dL) - Liver function marker
• CHE: Cholinesterase (U/L) - Liver enzyme
• CHOL: Cholesterol (mg/dL) - Lipid marker
• CREA: Creatinine (mg/dL) - Kidney function marker
• GGT: Gamma-glutamyltransferase (U/L) - Liver enzyme
• PROT: Total protein (g/dL) - Protein marker

Dataset Statistics

Category Distribution

Category	Count	Percentage
0=Blood Donor	452	73.5%
1=Hepatitis	73	11.9%
3=Cirrhosis	30	4.9%
2=Fibrosis	30	4.9%
0s=suspect Blood Donor	30	4.9%

Demographics

• Sex Distribution: 55.1% male, 44.9% female
• Age Range: 18-80 years
• Mean Age: 47.4 years

Age by Category

Category	Mean Age	Age Range
Blood Donor	44.5	32-77
Suspect Blood Donor	41.3	47-74
Hepatitis	44.1	19-61
Fibrosis	50.4	29-75
Cirrhosis	58.8	38-74

Usage

Loading the Dataset

import pandas as pd

# Load the dataset
df = pd.read_parquet('data/hepatitis_c.parquet')

# Display basic information
print(f"Dataset shape: {df.shape}")
print(f"Columns: {list(df.columns)}")

Data Exploration

# View category distribution
print("Category distribution:")
print(df['Category'].value_counts())

# View sex distribution
print("\nSex distribution:")
print(df['Sex'].value_counts())

# Summary statistics for laboratory values
print("\nLaboratory values summary:")
print(df[['ALB', 'ALP', 'ALT', 'AST', 'BIL', 'CHE', 'CHOL', 'CREA', 'GGT', 'PROT']].describe())

Machine Learning Applications

from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier

# Prepare features and target
feature_cols = ['Age', 'Sex', 'ALB', 'ALP', 'ALT', 'AST', 'BIL', 'CHE', 'CHOL', 'CREA', 'GGT', 'PROT']
X = df[feature_cols]
y = df['Category']

# Encode categorical variables
X = pd.get_dummies(X, columns=['Sex'], drop_first=True)

# Split the data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y)

# Train a model
model = RandomForestClassifier(random_state=42)
model.fit(X_train, y_train)

# Evaluate
accuracy = model.score(X_test, y_test)
print(f"Model accuracy: {accuracy:.3f}")

Data Types and Missing Values

• All features are numerical except Category and Sex
• No missing values in the dataset
• Category and Sex should be encoded for machine learning models
• Age is provided as integers (years)
• Laboratory values are floating-point numbers

Target Variable

The Category column serves as the classification target:

• Multi-class classification: 5 categories of liver health status
• Binary classification: Consider grouping (0, 0s vs 1, 2, 3) for healthy vs diseased
• Ordinal classification: Disease progression from healthy to cirrhosis

Feature Engineering Suggestions

• Demographic encoding: Convert Sex to binary (0/1)
• Age groups: Create age brackets if needed
• Laboratory ratios: Consider ALT/AST ratios, BIL/ALB ratios
• Feature scaling: Laboratory values may benefit from standardization

Performance Considerations

• Memory efficient: Parquet format provides fast loading and compression
• Stratified sampling: Use stratified cross-validation due to class imbalance
• Class weights: Consider balanced class weights for imbalanced categories