-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathfeatureConfig.yaml
More file actions
37 lines (29 loc) · 1.55 KB
/
featureConfig.yaml
File metadata and controls
37 lines (29 loc) · 1.55 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
# Which index components to include in the feature pool.
index:
GSPC: true # S&P 500
DJI: false # [WIP, Unsupported] Dow Jones Industrial Average
IXIC: false # [WIP, Unsupported] NASDAQ Composite
NDX: false # [WIP, Unsupported] NASDAQ-100
# Autoregressive feature generation settings.
lagSelection:
minLagsSelected: 2 # Minimum number of lags to select
maxLagsSelected: auto # [int, "auto"] Select based on results of lag significance tests. "auto" uses min(5, maxLag)
maxLag: 20 # Maximum lag to consider
selectionMethod: fdrAdjusted # [fdrAdjusted, rawPval]
sigLevel: 0.1
requireStability: true
stabilityFreq: 0.25 # Proportion of bootstrap samples in which a lag must be significant to be included
stabilityConfidence: 0.95 # Confidence level for stability check
stabilityCheckEvery: 25 # Check stability every N bootstrap samples
bootstrapSamples: 500 # Number of bootstrap samples to draw (Disable early stop to guarantee all samples are drawn)
minBootstrapSamples: 100 # Minimum number of bootstrap samples to draw before considering early stopping
earlyStop: true
blockLen: auto # [int, "auto"] Block length for bootstrap, "auto" uses max(5, floor(n ** (1 / 3))
rankTopN: 2 # If no lags are selected by tests, select the top N lags by p-value
hacBandwidth: auto # [int, "auto"] Bandwidth for HAC standard errors, "auto" uses Newey-West optimal
randomSeed: null # [int, null] Random seed for reproducibility, null for random
featurePool:
autoregressive: true
sentiment: true
sectoral: true
technical: true