# 基于Random Forest的決策策略 > 來源：https://uqer.io/community/share/54a10ef8f9f06c4bb886324b 版本：1.0 作者：李丞 聯系：cheng.li@datayes.com 利用隨機樹分類算法，通過歷史價格的上升狀態變化規律，預測下一日股價變動的方向。預測上漲則買入，下跌則賣出（如果可以的話）； ```py from sklearn.ensemble import RandomForestClassifier from collections import deque import pandas as pd import numpy as np start = pd.datetime(2010, 4, 1) end = pd.datetime(2014, 9, 16) longest_history = 1 bm = 'HS300' universe = ['600000.XSHG'] csvs = [] capital_base = 1e5 refresh_rate = 1 window_length = 10 def initialize(account): account.security = universe[0] account.window_length = window_length account.classifier = RandomForestClassifier() # 先進先出的deque序列，設定了最長的長度，在序列超過最長長度的時候，會將頭部序列移出 account.recent_prices = deque(maxlen=account.window_length+2) # 保存最近的股價 account.X = deque(maxlen=100) # 自變量 account.Y = deque(maxlen=100) # 應變量 account.prediction = 0 # 保存最近的預測值 def handle_data(account): hist = account.get_history(1) if account.security in hist: account.recent_prices.append(hist[account.security]['closePrice'][0]) # 更新最近的股價 if len(account.recent_prices) >= account.window_length+2: # 如果我們已經獲取了足夠的股價 RecentPrice=list(account.recent_prices) # 將deque轉換為對應的list # 制作一組1和0，標記股價是否相對于上一日價格上升。 changes = np.diff(RecentPrice) > 0 account.X.append(RecentPrice[1:-1]) account.Y.append(changes[-1]) if len(account.Y) >= 100: # 已經擁有足夠的數據im account.classifier.fit(account.X, account.Y) # 設定模型 account.prediction = account.classifier.predict(changes[1:]) # 預測 # 如果過大0.5，買入；小于0.5，賣出 if account.prediction > 0.5: buyAmount = int(account.position.cash / hist[account.security]['closePrice'][0]) order(account.security, buyAmount) else: order_to(account.security, 0) ```