for xsn in sn.sheet_names:datas = pd.read_excel('data.xlsx',sheet_name=xsn)m = list(datas.columns)[1:]tmp = pd.DataFrame({'date':pd.date_range(datas.iloc[0:,0].min(),datas.iloc[0:,0].max())})datas['date']=pd.to_datetime(datas['date'],format='%Y/%m/%d')datas = pd.merge(datas,tmp,on='date',how='outer').sort_values('date')# 线性插值for name in m:datas[name] = datas[name].interpolate(method='linear')# 对数据进行四舍五入for name in m:datas[name]=[round(value,2) for value in datas[name]]datas['volume']=[round(value) for value in datas['volume']]# 重新排列索引datas.index = range(len(datas))# 转换时间类型为objectdatas['date'] = [x.strftime('%Y/%m/%d') for x in datas['date']]datas.to_excel(writer,sheet_name=xsn,index=False)

for xsn in sn.sheet_names:# 索引值设置datas = datas.set_index("date")close_datas = datas['close']# 重采样resampled = close_datas.resample('m').ohlc()# 波动幅度,月份# (resampled.high-resampled.low)/resampled.lowripple = (resampled.high-resampled.low)/resampled.lowripple_datas.append(ripple.mean())# close_datas.plot(figsize=(8,6))# 收益率# 最大收益率 (close_datas.max() - close_datas.min())/close_datas.min()# 总收益率totol_growth = close_datas.iloc[-1]/close_datas.iloc[0]old_date = close_datas.index[0]new_date = close_datas.index[-1]# 复合收益率complex_growth.append(totol_growth**(1.0/(new_date.year-old_date.year))*100)# 重采样tmp_datas = datas['volume']resampled1 = tmp_datas.resample('m').last()# 成交量稳定性volume_datas.append(round(sum([(datas['volume'].iloc[i+1]-datas['volume'].iloc[i])/datas['volume'].iloc[i] for i in range(len(resampled1)-1)])/len(resampled1),3))

%% 正向化处理
[n,m] = size(datas_matrix);
% 正向化处理的数据所在列
Pos = [1,3];
% 指标类型：1：极小型，2：中间型，3：区间型
ch = [1,2];
% 循环处理每一列
for i = 1 : size(Pos,2)datas_matrix(:,Pos(i)) = Forward_processing(datas_matrix(:,Pos(i)),ch(i),Pos(i))
end
%% 矩阵标准化
datas_S_matrix = datas_matrix ./ repmat(sum(datas_matrix.*datas_matrix) .^ 0.5, n, 1)%%
model = ["abc001","abc002","abc003","abc004","abc005","abc006","abc007","abc008","abc009","abc010"];
%% 熵权法
p = datas_S_matrix./sum(datas_S_matrix);
k = 1/log(n);
r = zeros(n,m);
for i = 1:nfor j = 1:mif p(i,j) == 0r(i,j) = 0;elser(i,j) = log(p(i,j));endend
end
e = -k*sum(p.*r,1);
d = ones(1,m)-e
weight = d./sum(d)
score = sum(weight.*datas_S_matrix,2);
results1 = 0 + (100-0)/(max(score)-min(score)).*(score - min(score));
[sorted_score,index] = sort(results1 ,'descend');
rivers1 = [];
for i = 1:nrivers1 = [rivers1;model(index(i))];
end
s = [rivers1,sorted_score]

# 重采样，以周为单位
stock_week = datas['close'].resample('W-MON').mean()
stock_train = stock_week['2019':'2020']
# stock_train.plot(figsize=(12,8))
# plt.legend(bbox_to_anchor=(1.25,0.5))
# sns.despine()
# 一阶差分,平稳性较差
stock_diff=stock_train.diff()
stock_diff=stock_diff.dropna()# acf,确定q值=1
acf = plot_acf(stock_diff,lags=20)# pacf,确定p值=1
pacf = plot_pacf(stock_diff,lags=20)# 训练模型
model=ARIMA(stock_train,order=(1,1,1),freq='W-MON')
result=model.fit()
pred=result.predict('20191111','20200330',dynamic=True,typ='levels')
plt.figure(figsize=(6,6))
plt.xticks(rotation=45)
plt.plot(pred)
plt.plot(stock_train)