Trend-Following Strategies in Futures Markets

[Pages:35]Trend-Following Strategies in Futures Markets

A MACHINE LEARNING APPROACH

Art Paspanthong Divya Saini Joe Taglic

Raghav Tibrewala Will Vithayapalert

1

Outline

Overview of Data and Strategy Feature Generation Model Review

> Linear Regression > LSTM > Neural Network Portfolio Results Conclusion

2

Overview of Data and Strategy

3

Our Task

Problem Statement:

Replicate and improve on the basic ideas of trend following.

4

Datasets

Datasets of Commodities Futures

Energy Crude Oil

Metals Gold Silver Copper

Agriculture Corn Wheat

Soybean

Time Frame: 1 - 6 months Expiration Source: Quandl

Total Assets Considered 42 Different Assets

Filtered out by liquidity

Total Assets Traded 36 Different Assets

5

Data Exploration

Volatility Plot

Correlation Plot

6

Data Exploration

Gold 6-Month Futures

Sell Signal

Weak/False Signal

Buy Signal

7

Strategy

Traditional Trend Following

Not about prediction Involves quickly detecting a trend and riding it, all while managing when to exit at the right moment

Our Approach

Use traditional trend following indicators to predict returns with machine learning techniques Use a portfolio optimizer to weight assets using the predicted returns Adhere to traditional investment practice with stop-loss

8

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download