from tkinter import ttk, filedialog, messagebox
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from ngboost import NGBRegressor
from catboost import CatBoostRegressor
import lightgbm as lgb
import xgboost as xgb
from sklearn import metrics
import random
import time
# 定義平均模型類
class AverageModel:
def __init__(self, models, weights=None):
self.models = models
self.weights = weights if weights else [1/len(models)] * len(models)
def fit(self, X, y, X_val, y_val):
for model in self.models:
if isinstance(model, lgb.LGBMRegressor):
model.fit(X, y, eval_set=[(X_val, y_val)], eval_metric='rmse', callbacks=[lgb.early_stopping(stopping_rounds=100)])
elif isinstance(model, xgb.XGBRegressor):
model.fit(X, y, eval_set=[(X_val, y_val)], verbose=False)
elif isinstance(model, CatBoostRegressor):
model.fit(X, y, eval_set=(X_val, y_val), use_best_model=True, verbose=False)
elif isinstance(model, NGBRegressor):
model.fit(X, y, X_val=X_val, Y_val=y_val)
def predict(self, X):
predictions = []
for model in self.models:
predictions.append(model.predict(X))
return np.average(predictions, axis=0, weights=self.weights)
for i, model_name in enumerate(self.model_names):
tk.Label(self.root, text=f"{model_name}權重:").grid(row=8 + i, column=0, padx=10, pady=5)
entry = tk.Entry(self.root, state=tk.DISABLED)
entry.grid(row=8 + i, column=1, padx=10, pady=5)
self.model_entries[model_name] = entry
all_vars = [self.var_lgb, self.var_xgb, self.var_cat, self.var_ngb]
selected_vars = random.sample(all_vars, random.randint(1, len(all_vars)))
for var in selected_vars:
var.set(True)
def update_weight_entries(self, event):
weighting_method = self.combo_weighting.get()
if weighting_method == "自定義權重":
for model_name, entry in self.model_entries.items():
entry.config(state=tk.NORMAL)
else:
for model_name, entry in self.model_entries.items():
entry.delete(0, tk.END)
entry.config(state=tk.DISABLED)
def predict(self):
target = self.combo_target.get()
if not target or not self.filepath:
messagebox.showerror("錯誤", "請選擇一個目標變量并加載數據。")
return
X = self.df.drop([target], axis=1)
y = self.df[target]
self.models = []
selected_models = []
weights = []
if self.var_lgb.get():
self.models.append(model_lgb)
selected_models.append("LightGBM")
if self.var_xgb.get():
self.models.append(model_xgb)
selected_models.append("XGBoost")
if self.var_cat.get():
self.models.append(model_cat)
selected_models.append("CatBoost")
if self.var_ngb.get():
self.models.append(model_ngb)
selected_models.append("NGBoost")
if len(self.models) == 0:
messagebox.showerror("錯誤", "沒有選擇模型。")
return
if self.combo_weighting.get() == "自定義權重":
try:
weights = [float(self.model_entries[model_name].get()) for model_name in selected_models if self.model_entries[model_name].get()]
if len(weights) != len(self.models) or not np.isclose(sum(weights), 1.0):
raise ValueError
except ValueError:
messagebox.showerror("錯誤", "請提供有效的權重，并確保權重之和為1。")
return
if not weights:
weights = [1/len(self.models)] * len(self.models)
