Jainam-11/symptom-based-disease-prediction-v2

Symptom-Based Disease Prediction Dataset (Confidence-Aware) – v2 是一个改进和更结构化的医疗推理数据集，专为大型语言模型（LLMs）和AI驱动的医疗研究设计。该版本引入了更清洁的格式、改进的疾病分组、更好的置信度分离、增强的一致性和更适合微调的输出。每个样本以自然语言形式呈现症状，并根据置信度水平（高置信度、中置信度、低置信度）生成多个可能的疾病类别。数据集旨在支持医疗推理、多标签疾病预测、指令微调和医疗焦点AI实验。关键特征包括症状到疾病推理、多疾病预测支持、置信度感知的疾病分类、指令风格格式、处理重叠和模糊症状、JSONL格式和微调导向结构。数据集格式为JSON Lines，每个样本包含输入（症状描述）和输出（基于置信度的疾病列表）。示例输入为患者报告的症状，输出为按置信度分类的疾病。该数据集适用于医疗推理研究、LLM微调、医疗AI助手、症状分析系统、多标签分类研究、提示工程实验、临床NLP研究和教育AI应用。与版本1相比，版本2改进了格式一致性、置信度分离、疾病标签、指令响应对齐、JSON结构稳定性、LLM兼容性，减少了噪声输出，并提高了未来版本的可扩展性。数据集仅用于研究、实验、教育和AI开发目的，不能替代专业医疗建议、诊断或治疗。

Symptom-Based Disease Prediction Dataset (Confidence-Aware) – v2 is an improved and more structured medical reasoning dataset designed for Large Language Models (LLMs) and AI-driven healthcare research. This version introduces cleaner formatting, improved disease grouping, better confidence separation, enhanced consistency, and more fine-tuning-friendly outputs. Each sample presents symptoms in natural language form and generates multiple possible diseases categorized by confidence levels (high confidence, medium confidence, low confidence). The dataset is designed to support medical reasoning, multi-label disease prediction, instruction tuning, and healthcare-focused AI experimentation. Key features include symptom-to-disease reasoning, multi-disease prediction support, confidence-aware disease categorization, instruction-style formatting, handling overlapping and ambiguous symptoms, JSONL format, and fine-tuning-oriented structure. The dataset is stored in JSON Lines format, with each sample containing an input (symptom description) and an output (confidence-based disease lists). Example input includes patient-reported symptoms, and output includes diseases classified by confidence. It is intended for use in medical reasoning research, LLM fine-tuning, healthcare AI assistants, symptom analysis systems, multi-label classification research, prompt engineering experiments, clinical NLP research, and educational AI applications. Compared to Version 1, Version 2 improves formatting consistency, confidence separation, disease labeling, instruction-response alignment, JSON structure stability, LLM compatibility, reduces noisy outputs, and enhances scalability for future versions. The dataset is strictly for research, experimentation, educational purposes, and AI development, and must not be used as a replacement for professional medical advice, diagnosis, or treatment.