Fig 1.
Overview of the PRML model.
Fig 2.
Main pattern recognition algorithm.
Fig 3.
Description of a candlestick and candlestick patterns.
Table 1.
Parameters of the four machine learning models used in prediction schedule.
Fig 4.
Flowchart of investment strategy construction for two-day patterns and three-day patterns.
Table 2.
Data samples used in the machine learning models.
Fig 5.
Accuracy overviews of PRML and pure machine learning models.
Fig 6.
The number of the four machine learning methods supporting highest prediction accuracy.
Fig 7.
Portfolio forecasting performance of 1, 2, 3, 5, 7, 10 days ahead: PRML vs. ML.
Table 3.
Finance performance of forecasting one day ahead: PRML vs. ML.
Fig 8.
Portfolio return performance of two-day patterns predicting 1, 2, 3, 5, 7, 10 days ahead.
Table 4.
Finance performance of two-day patterns predicting one day ahead.
Fig 9.
Portfolio return performance of three-day patterns predicting 1, 2, 3, 5, 7, 10 days ahead.
Table 5.
Finance performance of three-day patterns predicting one day ahead.
Fig 10.
One day ahead forecasting performance of MLP and LSTM.
Table 6.
One day ahead performance of MLP and LSTM.
Table 7.
Finance performance of sliding windows in two-day patterns predicting one day ahead.
Fig 11.
Portfolio return of two-day predictions one day ahead with 0.2% transaction cost.
Table 8.
Finance performance of two-day predictions one day ahead with 0.2% transaction cost.