Logic Strategy machine learning to identify different candlesticks
The logic strategy uses various aspects of artificial intelligence, genetic algorithms and neural networks to identify candlesticks from price data. Here we will be fully transparent on how the Logic strategy uses price data to dynamically identify different candle
Step 1: Break each candlestick into 7 ratios
Firstly, the logic strategy identifies 4 inferred variables:
Body size = Open - Close
Upper wick length = High - Top of "Body size"
Lower wick length = Low - Bottom of "Body size"
Total size of candle = High - Low
Now with the 4 inferred variables the Logic Strategy looks for a specific candle type from profession recognition, this varies from candle to candle. What it is primarily searching for is the ratios between the "4 inferred variables".
The 7 ratios are:
Body size to Upper wick length,
Body size to Lower wick length
Body size to Total size of candle
Upper wick length to Lower wick length
Upper wick length to Total size of candle
Lower wick length to Total size of candle
Whether the candle is bullish or bearish
Step 2: Build a training set using a professional that identified X type of candles (shooting star)
This step requires a qualified financial engineer, such as a logikfx employee. To collect a large list of candlesticks which are shooting stars (candlestick type). The data must include price data: open, high, low and close.
The 1 in the 5th column means it is a specified candle type, if it was 0 it would mean it is not a specified candle and it doesn't meet the criteria of the ratios.
Once the Logic Strategy acquires these values, it can now start learning what a certain candlestick type is.
To make sure the Artificial intelligence is identifying the candlestick type correctly we include candles which are not shooting stars.
These are further down the list which will be marked as 0 in the far right column.
Step 3: Numerical Identification of Candlestick
This step is where the logic strategy uses the numbers from above to work out what the candle will look like. This is the same way a professional would identify a candlestick type visually.
Once the logic strategy knows these types of variables and values are a shooting star candlestick it can now identify them in the future, how a professional would see them on the chart. This is much better than hard coded candlesticks because hard code struggles to differentiate between different candlesticks.
Using machine learning, neural networks and genetic algorithms the logic strategy can dynamically identify candlesticks automatically to be used in analysis. If there is a candlestick we do not believe should be in the identification then we add that to the training set list and the artificial intelligence will learn to remove it from the classification.
To keep up to date with education, analysis and research head to our info centre.