MLTA

Machine Learning Driven Technical Analysis Library in Python.

Motivation

Technical Analysis has long been a stable for generating insights for trading. While the predictive power of each individual indicator is up for a debate, it is no denying that technical indicators are useful tools to create summary statistics for time-series data. This usefulness is not limited to finance, but rather any domain with necessity to deal with time series.

However, most indicators are interpreted in a too simplistic way. For example, candlestick patterns classify each candle into one of a couple classes. However, we would get more useful information if we can also measure how close a candle matches each class.

MLTA enables this possibility. Instead of returning classes, MLTA generates the likelihood of a candle belonging to a given class. The result is more granula information which leads to an improvement in accuracy.

Installation

MLTA is not yet on pypi. You can install the package through the following command:

pip install git+https://github.com/palmbook/MLTA.git

Usage

Candlestick Patterns

from MLTA import Candlestick

candle_model = Candlestick()

# df must have columns open, high, low, and close (in lower letters)
# class_prob contains columns of probabilities for each possible class
class_prob = candle_model.CDL3INSIDE(df)

Available methods are

• CDL3INSIDE
• CDL3LINESTRIKE
• CDL3OUTSIDE
• CDL3WHITESOLDIERS
• CDLCLOSINGMARUBOZU
• CDLCOUNTERATTACK
• CDLDOJI
• CDLDRAGONFLYDOJI
• CDLGAPSIDESIDEWHITE
• CDLGRAVESTONEDOJI
• CDLHAMMER
• CDLHARAMI
• CDLHOMINGPIGEON
• CDLINVERTEDHAMMER
• CDLLADDERBOTTOM
• CDLLONGLEGGEDDOJI
• CDLLONGLINE
• CDLMARUBOZU
• CDLMATCHINGLOW
• CDLMORNINGDOJISTAR
• CDLMORNINGSTAR
• CDLRICKSHAWMAN
• CDLRISEFALL3METHODS
• CDLSEPARATINGLINES
• CDLSHORTLINE
• CDLSTICKSANDWICH
• CDLTAKURI
• CDLTASUKIGAP
• CDLUNIQUE3RIVER
• bullishPin
• bearishPin