数据挖掘大作业-基于python的电信客户流失预测与分析

import torch import torch.nn as nn from matplotlib import pyplot as plt import os from Dataset.Dataset import csv_dataset from module.model import MLPClassifier import torch.nn.functional as F from torch.utils.data import Dataset from torch.utils.data import DataLoader from torch.utils.data import random_split import numpy as np import pandas as pd import tqdm from sklearn.model_selection import StratifiedKFold os.environ['KMP_DUPLICATE_LIB_OK'] = 'TRUE' device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") print('[Info]: use:', device) Data_dir = './Datas/Datas.csv' label_dir = './Datas/labels.csv' np_file_dir = './Datas/random_allnums.npy' SAVEDIR = './runs' ALL_ACC, ALL_SEN, ALL_PRE = 0, 0, 0 for K_flod in range(1, 6): train_set = csv_dataset(Data_dir, label_dir, k=K_flod, train=True, path_np=np_file_dir) test_set = csv_dataset(Data_dir, label_dir, k=K_flod, train=False, path_np=np_file_dir) train_loader = DataLoader(dataset=train_set, batch_size=8, shuffle=True, ) test_loader = DataLoader(dataset=test_set, batch_size=8, shuffle=True, ) model = MLPClassifier() model = model.to(device) weights = torch.FloatTensor([2]) criterion = nn.BCEWithLogitsLoss(pos_weight=weights).cuda() optimizer = torch.optim.Adam(model.parameters(), lr=0.0003) loss_list = [] TP, TN, FP, FN = 0, 0, 0, 0 for epoch in range(15): bar = tqdm.tqdm(train_loader) for X, y in bar: bar.set_description("epoch: %s" % str(epoch)) X = X.to(device, dtype=torch.float32) y = y.to(device, dtype=torch.float32) y_pred = model(X) loss = criterion(y_pred, y) optimizer.zero_grad() loss.backward() optimizer.step() loss_list.append(loss.data.item()) bar.set_postfix(loss=loss.data.item()) torch.save(model.state_dict(), os.path.join(SAVEDIR, str(epoch)) + '.pt') model.eval() bar2 = tqdm.tqdm(test_loader) for X, y in bar2: X = X.to(device, dtype=torch.float32) # y = y.to(device, dtype=torch.float32) y_pred = model(X) y_pred = F.sigmoid(y_pred) y_pred[y_pred > 0.5] = 1 y_pred[y_pred <= 0.5] = 0 y_pred = torch.Tensor.cpu(y_pred) sTP = np.array(((y == 1) & (y_pred == 1)).sum()) sFN = np.array(((y == 1) & (y_pred == 0)).sum()) sTN = np.array(((y == 0) & (y_pred == 0)).sum()) sFP = np.array(((y == 0) & (y_pred == 1)).sum()) TP = sTP + TP TN = sTN + TN FP = sFP + FP FN = sFN + FN print('TP:',TP) print('TN:',TN) print('FP:',FP) print('FN:',FN) acc = (TP + TN) / (TP + TN + FP + FN) sen = (TP) / (FN + TP) pre = (TP) / (TP + FP) print('flod:' + str(K_flod) + ' acc:', acc) print('flod:' + str(K_flod) + ' sen:', sen) print('flod:' + str(K_flod) + ' pre:', pre) ALL_ACC = ALL_ACC + acc ALL_SEN = ALL_SEN + sen ALL_PRE = ALL_PRE + pre plt.plot(np.linspace(0, 100, len(loss_list)), loss_list) plt.show() print('Flod:',K_flod,' Finish!') print('==========================') print('ALL_ACC:', ALL_ACC / 5) print('ALL_SEN:', ALL_SEN / 5) print('ALL_PRE:', ALL_PRE / 5)