Price prediction platform for fund in Chinese market.

This platform provides a suite of tools to help users achieve following goals. The whole platform could be divided into 3 function parts

Function parts

1. fund data operator
   • get fund data both historical and the latest.
   • automatically parsing fund data and save to db
   • db file/type can be set in config.json
2. stragegy management
   • this platform could be connected to a flask web app
   • strategy could be set in config file.
   • strategy verification platform/tools
3. asset management
   • personal virual cash/asset management, multi-users supported.
4. simulate purchasing/selling fund operations.
   • simulate purchasing/selling fund
   • operation record is kept in db.

Project architecture introduction

utils

This package provides a suite of tool including scapping fund data, clearing data, operating data and wrapped CRUD operations.

1. get_fund.py -> Data_operator class defined a set of functions to scraping data from eastmoney

   • get_one_fund() -> scrapping one fund at a certain date slot
   • read_multiple() -> a subprocess called in multi-thread/multi-process scrapping
   • get_funds_multi_process() -> accelerate scrapping data by multi-processes.
   • get_funds_multi_thread() -> accelerate scrapping data by multi-threads.
   • get_funds() -> scrapping fund by single process/single thread.
   • load_fund() -> select target fund data from database.
   • load_funds() -> select multiple target funds data from database.
   • load_asset() -> laod user's asset from db.
   • load_record() -> load purchasing/selling record data from db.
   • update_fund() -> update the fund data,crud on fund in db.
   • get_fund_list() -> read fund list.
   • get_realtime_price() -> get the real time fund price.

2. operatefund.py -> Operator class defines a simulation of buy/sell operations.

   • buy_at_date() -> purchase a certain amount of fund at the price of a specific date.
   • sell_at_date() -> sell a certain amount of fund at the price of a specific date.
   • sell_fund_all() -> sell all fund shares hold in hand.
   • get_asset_value() -> calculate value of fund hold in hand.

3. readconfg.py -> read config file.

4. playground.py -> for unit test and trials.

5. wechat.py -> for future wechat notification function.

strategies

Some samples of strategies, jupyter note or stragegy py file.

1. fixedrule.py -> define some fixed rules to decide when to buy/sell.
2. SingleLSTM.ipynb -> LSTM model on predicting one fund price.
3. verify.py -> back test/verify the strategy.
4. AffinityCluster.ipynb -> Apply cluster algorithm on funds.

models

Define database schema and basice data operations.

1. Fund
   • fund_code
   • date
   • price
   • accumulate
   • daily_rate
   • pruchase_state
   • ransom_state
   • dividends
2. User
   • userid
   • password
   • username
   • my_cash
3. Record
   • userid
   • fund_code
   • date
   • price
   • accumulate
   • units
   • buy_sell
4. Asset
   • userid
   • my_fund
   • my_units
   • my_cost

frontend

User asset/Record management page, under consturction. vue.js.

backend

flask supported

others

config.json

platform for analyst to create algorithm in strategies

write your strategies in ./strategies and config it in config.json

please check the requirements.txt to make sure your env is right

Recomand you to use conda to install or virtualenv is also a good choice.