Dear People, Dear Stephan
I’m doing data mining on several time frames 30 min, daily, weekly (plan is also to go to 5min and 1min) using data (and indicators) from tradingview. I want to analyze first NQ, and then other futures to look for correlations. Current focus is on “intraday technical analysis” (with a twist derived from weekly and daily biases).
My target dataframes (30min, daily, weekly) are as follow:
Date
Time (if frequency below daily)
OHLC
Volume
Return 1 2 3 5 8 11 periods (on close, high, low - in ticks, not %)
Factors are (as said) purely from technical analysis. Below a short sample
0 derived: weekday, time of day, month
1 Range, Range-categorized 1-5 (narrow - wide)
2 Bar2type: up, down, inside, outside (compare with previous bar)
3 ValueArea2type: up, up-overlapping, inside, outside
4 closeStreak: trending indicator, if cl >= cl1 and closeStreak>0 then closeStreak = closeStreak + 1 elseif cl <= cl1 and closeStreak<0 closeStreak = closeStreak - 1 else initializes to 1 or -1 when closing trend changes
5 Tail indicator 1 bullish, 0 no tail, -1 bearish tail (o & c away from high or low)
6 “trading indicators”: if price goes from top of value area to bottom and then closes inside the value area, should be a buy (80 Percent Rule Strategy | PDF | Market Trend | Market (Economics))
And a big etcetera related to the technical analysis books, like Kaufman, James F. Dalton, Larry Williams, Linda Bradford Rashke, Wyckoff. There are many.
Idea is to create a dataframe with weekly, daily and intraday indicators and feed ML. Many of the factors I want to mine are not orthogonal (2 & 3 by example). One of the aims of the study is to discard weaker factors / classify.
- Which ML approach can be recommended for such a study? By example a Regression can be a good fit? Or is better to go Deep Learning? (or whatever else)
- Anyone has tried a similar or comparable approach?
Thanks for the attention
Luis