Notebooks and Code for ML based quant strategies and research.
Original data distribution, prediction by a mixture distribution implemented using tensorflow probability(TFP) and the prediction by sklearn Gaussian Mixture Model.
TFP mixture model components
Predicted trends by the mixture model
Here we sample and plot 'close' value based on the volume. (Close value at every "xx" volume)
Here we generate trends (% increase or decrease in SPY) and find the trend strength (value between 0 and 1 depending on how long the trend last) and trend period (how long the trend last).
This notebook has a method to find the best EMA lines that represent 'uptrends' and 'downtrends' in S&P 500.
This shows how to use probabilistic Logistic Regression to detect outliers in a dataset. Prob LR is useful to guage the confidence of a decision. This is helpful to decide the risk of a bet.