Predict profitability of trades based on indicator buy / sell signals

Trade profitability analysis for trades based on various indicators signals:

MACD

Simple Moving Average

Exponential Moving Average

Trading assumptions:

Trade is profitable if, profit >0
Buy / sell happen the following day of the signal
Buy / sell are taken 10% from the open price towards close price

Machine learning assumptions:

Binary classification: 1 - profit, 0 - loss
A separate model for each company / ticker
Model is trained vs optimal precision

Machine learning models used:

Linear Support Vector Classifier
Decision Tree Classifier
Random Forest Classifier
Gradient Boosting Classifier
XGBoost Classifier
Keras classifier

Trade analysis intermediate results:
30-40% of trades based on indicator signals are profitable
In general trades on SMA signals are more often profitable than the ones based on EMA signals

Trade profitability predictions intermediate results (based on test data)/
The precision of the predictions is oscilating around 70%, which is pretty good, considering that the analysts estimate other signals accuracy as 30 to 50% (double top, shoulder & arms, etc). This means, there is ~70% chance that predicted trade will be profitable (Reminder: profitable -> profit > 0)
However, the recall is only around 15%, which means that very the model pick-up very few of the actually profitable trades.

#Detailed analysis tbc

Predict profitability of trades based on indicator buy / sell signals

Related tags

Overview

Predict profitability of trades based on indicator buy / sell signals

Owner

Tomasz Porzycki

Machine learning template for projects based on sklearn library.

Credit Card Fraud Detection, used the credit card fraud dataset from Kaggle

ETNA is an easy-to-use time series forecasting framework.

A collection of interactive machine-learning experiments: 🏋️models training + 🎨models demo

Practical Time-Series Analysis, published by Packt

DirectML is a high-performance, hardware-accelerated DirectX 12 library for machine learning.

An open-source library of algorithms to analyse time series in GPU and CPU.

Combines Bayesian analyses from many datasets.

A demo project to elaborate how Machine Learn Models are deployed on production using Flask API

ArviZ is a Python package for exploratory analysis of Bayesian models

Distributed deep learning on Hadoop and Spark clusters.

决策树分类与回归模型的实现和可视化

LibTraffic is a unified, flexible and comprehensive traffic prediction library based on PyTorch

Sequence learning toolkit for Python

GAM timeseries modeling with auto-changepoint detection. Inspired by Facebook Prophet and implemented in PyMC3

This is a Cricket Score Predictor that predicts the first innings score of a T20 Cricket match using Machine Learning

Pydantic based mock data generation

monolish: MONOlithic Liner equation Solvers for Highly-parallel architecture

A high performance and generic framework for distributed DNN training

pandas, scikit-learn, xgboost and seaborn integration