0

我需要将自己的数据用于滑索项目。每当我尝试时，我都会收到此错误：

/Library/Python/2.7/site-packages/zipline/sources/data_source.pyc in <dictcomp>((target, (mapping_func, source_key)))
     47         """
     48         row = {target: mapping_func(raw_row[source_key])
---> 49                for target, (mapping_func, source_key)
     50                in self.mapping.items()}
     51         row.update({'source_id': self.get_hash()})

ValueError: cannot convert float NaN to integer

这是我正在运行的交易算法：

from zipline.algorithm import TradingAlgorithm
from zipline.api import order_target, order, record, symbol, history, add_history
import numpy as np
from pandas import Series, DataFrame, Panel
import pandas as pd

# Define algorithm
def initialize(context):
    context.dateIndex = 0

def handle_data(context, data):
    today = data.major_axis[context.dateIndex]

    if today > data.US9663871021[data.US9663871021.close.notnull()].index[0] and today < data.US9663871021[data.US9663871021.close.notnull()].last_valid_index():
        order(symbol('US9663871021'), 10)
        record(US9663871021=data[symbol('US9663871021')].price)
    if today > data.US7954351067[data.US7954351067.close.notnull()].index[0] and today < data.US7954351067[data.US7954351067.close.notnull()].last_valid_index():
        order(symbol('US7954351067'), 10)
        record(US7954351067=data[symbol('US7954351067')].price)

    if today == data.US9663871021[data.US9663871021.close.notnull()].last_valid_index():
        order_target(symbol('US9663871021'), 0)
        record(US9663871021=data[symbol('US9663871021')].price)
    if today == data.US7954351067[data.US7954351067.close.notnull()].last_valid_index():
        order_target(symbol('US7954351067'), 0)
        record(US9663871021=data[symbol('US7954351067')].price)    

    context.dateIndex = context.dateIndex + 1

def prepDf(fileName):
    df = pd.io.parsers.read_csv(fileName, index_col=[0],parse_dates=[0], na_values=["#N/A N/A"],
                            names=["date", "open","high","low","close","volume","mkt_cap"])
    df["price"] = df.close
    df.index = df.index.tz_localize('UTC')
    df = df[df.close.notnull()]
    return df

fileName = #fill in file name
fileName2 = #fill in file name

dictionaryOfDfs = {"US9663871021" : prepDf(fileName), "US7954351067": prepDf(fileName2)}

data = Panel(dictionaryOfDfs)
algo_obj = TradingAlgorithm(initialize=initialize, 
                            handle_data=handle_data)

# Run algorithm
perf_manual = algo_obj.run(data)

这个想法是我在数据应该是非 NaN 时买入，并在系列结束前卖出头寸。除此之外，应该不需要数据，但 zipline 坚持认为 NaN 会导致错误，即使不应该使用该值。

4

1 回答 1

0

经过研究，我认为解决方案是重新索引底层数据帧：

df1 = df1.reindex(index=data.major_axis, fill_value=0)
df2 = df2.reindex(index=data.major_axis, fill_value=0)

数据是熊猫面板

于 2015-05-11T17:05:53.403 回答