Add files via upload

client_list/client_list.py

@@ -0,0 +1,85 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Wed Feb 20 14:40:59 2019
+
+@author: howard
+"""
+
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Mon Feb 18 16:17:15 2019
+
+@author: howard
+"""
+
+import pandas as pd
+from xgboost import XGBClassifier
+from sklearn.model_selection import train_test_split
+import numpy as np
+
+def detect_str_columns(data):
+    '''
+    1. 偵測有字串的欄位
+    2. 挑選出來，準備encoding
+    
+    '''
+    strlist = list(set(np.where((data.applymap(type)==str))[1].tolist()))
+    return data.columns[strlist].tolist()
+
+def get_dummies(dummy, dataset):
+    ''''
+    make variables dummies
+    ref：http://blog.csdn.net/weiwei9363/article/details/78255210
+    '''
+    dummy_fields = list(dummy)
+    for each in dummy_fields:
+        dummies = pd.get_dummies( dataset.loc[:, each], prefix=each ) 
+        dataset = pd.concat( [dataset, dummies], axis = 1 )
+
+    fields_to_drop = dummy_fields
+    dataset = dataset.drop( fields_to_drop, axis = 1 )
+    return dataset
+
+# 讀取marketing資料
+data = pd.read_csv('marketing2.csv')
+
+
+# 偵測有字串的欄位
+str_columns = detect_str_columns(data.drop(columns = 'UID'))
+dataset = get_dummies(str_columns, data)
+
+# 確認全部都是數字 float, int, uint --> ML
+dataset.info()
+
+
+# 切分資料集
+X =dataset.drop(columns=['買A商品'])
+y =dataset['買A商品']
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 0) 
+
+# 保留UID
+train_uid = X_train['UID']
+test_uid = X_test['UID']
+
+# 3. -------------ML預測-------------
+
+# 設定xgb 分類模型
+del X_train['UID']
+del  X_test['UID']
+clf = XGBClassifier(n_estimators=300 ,random_state = 0, nthread = 8, learning_rate=0.5)
+model_xgb = clf.fit(X_train, y_train, verbose=True,eval_set=[(X_train, y_train), (X_test, y_test)])
+
+# 進行預測
+y_pred = model_xgb.predict(X_test)
+y_pred_prob = model_xgb.predict_proba(X_test)[:,1]
+
+# 精準客戶名單
+XGBClassifier_test_df=pd.DataFrame(y_test.values ,columns =['客戶對A商品【實際】購買狀態'])
+XGBClassifier_test_df['客戶對A商品【預測】購買機率'] = y_pred_prob
+test_uid = test_uid.reset_index().drop(columns = ['index'])
+XGBClassifier_test_df = pd.concat([test_uid,XGBClassifier_test_df], axis = 1)
+
+
+