Update app.py

app.py

@@ -1,104 +1,47 @@
-import os
 import gradio as gr
 import xgboost as xgb
 import numpy as np
 import pandas as pd
-from transformers import pipeline
 from huggingface_hub import hf_hub_download
 
-# **📌 先运行 `preprocess.py` 处理数据**
-processed_data_path = "processed_data.csv"
+# **📥 从 Hugging Face 下载 XGBoost 模型**
+repo_id = "YDluffy/lottery_prediction"
+model_filename = "lottery_xgboost_model.ubj"
+model_path = hf_hub_download(repo_id=repo_id, filename=model_filename)
 
-if not os.path.exists(processed_data_path):
-    print("📌 运行 `preprocess.py` 进行数据处理...")
-    os.system("python preprocess.py")
-
-# **📌 加载处理后的数据**
-df = pd.read_csv(processed_data_path)
-
-# **📌 加载 Hugging Face GPT-Neo 作为 LLM**
-print("📌 正在加载 GPT-Neo 模型...")
-generator = pipeline('text-generation', model='EleutherAI/gpt-neo-2.7B')
-print("✅ GPT-Neo 模型加载成功！")
-
-# **📌 加载 XGBoost 预测模型**
-model_path = hf_hub_download(repo_id="YDluffy/lottery_prediction", filename="lottery_xgboost_model.json")
-model = xgb.XGBRegressor()
+# **✅ 加载 XGBoost 预测模型**
+model = xgb.Booster()
 model.load_model(model_path)
 
-# **📌 预测函数**
-def predict_lottery(year, period, num1, num2, num3, num4, num5, num6, special):
-    # **从历史数据查找对应的月份和日期**
-    history = df[(df['期号_年份'] == year) & (df['期数'] == period)]
-    if not history.empty:
-        month = history.iloc[0]['月份']
-        day = history.iloc[0]['日期']
-    else:
-        month, day = 1, 1  # 默认值
-
-    # **计算中奖号码均值**
-    avg_number = np.mean([num1, num2, num3, num4, num5, num6])
-
-    # **形成特征数组**
-    features = np.array([[year, period, month, day, num1, num2, num3, num4, num5, num6, special, avg_number]])
-
-    # **进行预测**
-    prediction = model.predict(features)
+# **📥 读取历史开奖记录**
+history_data_path = "Mark_Six.csv"
+history_data = pd.read_csv(history_data_path)
 
-    # **🚀 修正浮点数问题：四舍五入为整数**
-    prediction = np.round(prediction).astype(int)
-
-    return prediction.tolist()
-
-# **📌 LLM 解析用户输入并调用 XGBoost**
-def chat_with_llm(user_input):
-    prompt = (
-        "请从以下问题中提取出预测所需的参数，并调用 XGBoost 预测模型。\n"
-        "返回格式：\n"
-        "年份:2025, 期号:16, 号码:[5,12,23,34,45,56], 特别号码:7\n"
-        "请勿返回额外信息，仅输出这个格式的数据。\n"
-        f"输入问题: {user_input}"
-    )
-
-    try:
-        response = generator(
-            prompt,
-            max_new_tokens=50,
-            temperature=0.1,
-            num_return_sequences=1
-        )
-        extracted_text = response[0]['generated_text']
-
-        # **🚀 确保 UTF-8 编码，移除特殊字符**
-        extracted_text = extracted_text.encode("utf-8", "ignore").decode("utf-8").strip()
-
-        # **✅ 确保返回格式正确**
-        if not all(keyword in extracted_text for keyword in ["年份:", "期号:", "号码:", "特别号码:"]):
-            return f"❌ LLM 解析失败：返回数据格式不正确。\n📢 LLM 解析结果: {extracted_text}"
-
-        # **解析 GPT-Neo 生成的文本**
-        parts = extracted_text.split(",")
-        year = int(parts[0].split(":")[1].strip())
-        period = int(parts[1].split(":")[1].strip())
-        nums = [int(x) for x in parts[2].split(":")[1].strip("[]").split()]
-        special = int(parts[3].split(":")[1].strip())
+# **📌 预测函数**
+def predict_lottery(year, period):
+    # **📌 1. 查找往年相同期数的开奖记录**
+    historical_matches = history_data[history_data["期数"] == period]
 
-        # **调用 XGBoost 进行预测**
-        prediction = predict_lottery(year, period, *nums, special)
+    # **📌 2. 计算历史趋势（如高频号码、奇偶比例）**
+    most_frequent_numbers = historical_matches.iloc[:, -7:].mode().iloc[0].tolist()
+    
+    # **📌 3. 生成 XGBoost 预测输入**
+    test_features = np.array([[year, period] + most_frequent_numbers])
+    dtest = xgb.DMatrix(test_features)
 
-        return f"📊 预测的号码是: {prediction}\n\n📢 LLM 解析的特征：{extracted_text}"
+    # **📌 4. 进行预测**
+    prediction = model.predict(dtest)
+    final_prediction = np.round(prediction).astype(int).tolist()
 
-    except Exception as e:
-        return f"❌ LLM 解析数据失败: {str(e)}\n📢 LLM 解析结果: {extracted_text}"
+    return final_prediction
 
-# **📌 Gradio Web 界面**
+# **📌 创建 API 接口**
 iface = gr.Interface(
-    fn=chat_with_llm,
-    inputs=gr.Textbox(label="请输入问题或期号信息"),
+    fn=predict_lottery,
+    inputs=["number", "number"],
     outputs="text",
-    title="六合彩预测模型",
-    description="GPT 解析输入信息，并调用 XGBoost 进行预测"
+    title="六合彩智能预测 API",
+    description="输入年份和期数，自动分析历史开奖记录，预测开奖号码"
 )
 
-# **📌 启动 Gradio 应用**
 iface.launch(share=True)