msbd5013_project

msbd5013 group project repository

Folder Usage

• data

Contains both format1 & format2 data from dropbox.

• analysis

Historical data analysis should be doing in this folder.

• model

Model building should be done in this folder.

• submissions

All submitted history project result archieve.

• others

Other support and helpful files or pages should be placed here.

File Usage

• backTest.py

Enterance for porgramming running.

(Data file directory have already be setted as './data/')

• backTest.ipynb

backTest in jupyter-notebook format.

(Also contains some debug code for strategy.)

• strategy.py

Strategy code.

• auxiliary.py

Support functions for stategy.

Helpful Links.

• about time series analysis : https://www.cnblogs.com/foley/p/5582358.html

• some source code about strategy and model: https://www.joinquant.com

• others/Analysis of Financial Time Series.pdf

• others/Time series prediction using dynamic Bayesian network.pdf

• LSTM-Neural-Network-for-Time-Series-Prediction

Meeting Summary

2018/10/04 会议结论

1. Strategy

   1.1 完善Cash_balance监控，要分清long，short时候对position的影响及cash的影响，保证不被平仓。

   1.2 急涨急跌的应急处理。细节完善

   1.3 联系TA关于清仓后，cash高于10000后，应该依然可以做交易的问题

2. Model

   2.1 基于Bayesian Network的初步建模。

   输入(t-1)时刻四个货币的价格, 输出为(t))时刻四种货币价格或者走势的预测。

3. Statistics & Analysis

   3.1 多个货币之间的走势联系和影响，分析及利用。

   合适的预测粒度（多少小时、或者天），以支持模型训练及优化

   走势“领头”数字货币，支持策略方面实时计算决策。

   3.2 交易量变化分析 - volume & price 走势关联分析

   3.3 急涨急跌后的走势分析，辅助决策。

4. Arragement

   Strategy - ZHANG Xichen

   S & A - GAO Han

   Model - WANG Shen, LING Liyang (coding support)

5. Deadline

   2018/10/18 (Week.3)

2018/10/11 会议结论

1. Lask week rank: No.1

2. Model

   2.1 Linear Regression.

   model/model_linear_regression.ipynb

   Y = (W.T)*X, where X is the data at time (T-n) of all 4 assests, Y is the predict assest price at time T.

   2.2 Logistic Regression.

   modify based on linear regression model

3. Strategy

   based on model, if the predict is trustable. trade.

4. Arragement

   1. Linear Regression parameter extend to (T-1), (T-2) ... (T-n) - ZHANG Xichen

   2. Logistic Regression Modification. - GAO Han

   3. Linear Regression parameter extend to volume. - LING Liyang

   4. Textbook method attempt. - WANG Shen

5. DDL

   2018/11/01 (Week.5)