LinearTree

Here, we show model performance in predicting the bitcoin price and their hyper-parameters.

Prediction

For scoring, Root Mean Squared Error was used.

Mean across all targets: 1,018.73

D+7	D+14	D+21	D+28
1,013.33	859.43	2,116.25	85.89

Mean across all targets: 13,704.62

D+7	D+14	D+21	D+28
2,501.42	16,099.47	19,147.52	17,070.09

Below are choices made for the model setup.

(Determined by Forward Selection and heuristically based on the model's assumptions)

param	description	value
include_indices	Whether to include market indices.	True
include_commodities	Whether to include commodities.	False
include_currencies	Whether to include currencies.	False

(Determined by Grid Search)

param	description	value
ohlc_ratio	Whether to add OHLC ratios of bitcoin price to the data.	True
lags	List of lags to use. None, if no lags are used.	None

(Determined by heuristically, considering the model's assumptions)

param	description	value
log	Whether to take the log of the data.	False
diff	Whether to difference the data.	True
normalize	Whether to normalize the data.	True

(Determined by Grid Search)

param	description	value
num_leaves	The number of leaves in the tree. A smaller number of leaves may reduce overfitting.	10
max_depth	The maximum depth of the tree. A negative value means no limit.	-1
min_data_in_leaf	The minimum number of data points required in a leaf. This is a regularization parameter.	5
num_boost_round	The number of boosting rounds, or the number of trees in the ensemble.	1