Time Series Forecast With Stock's Price Candlestick Patterns And Sequence Similarities

This study employs candlestick patterns and sequence similarities to forecast trends in time-series stock data. Time-series financial forecasting is a crucial component in risk management and portfolio optimization. Finding a suitable technique for financial engineering poses a challenge since it must meet the criteria of being transparent, robust, and compatible. There is more and more multi-modal data being used in research, and prediction algorithms are getting harder to understand and making it harder for data to move smoothly. Candlestick methods are expected to be effective in the scenarios above. The utilization of this tool is prevalent within financial markets due to its accessibility, comprehensibility, and acceptance within the technical analysis community. This study examines the initial and subsequent stages of forecasting. The process of sequential pattern mining involves the extraction of candlestick patterns from candlestick data with several dimensions. Correlation analysis then assesses the connection between these patterns and potential future trends. Furthermore, this study introduces a novel approach to identifying commonalities in candlestick charts that exhibit considerable differences. Two experimental groups were utilized to evaluate the technique's efficacy, employing a sample size of 800 equities from the Chinese stock market. The stability and accuracy of our approach surpass those of the SVM (50.83% and 51.32%) and LSTM (50.71% and 50.68%) models, exhibiting an average accuracy of 52.12% and 53.13%. This study can potentially contribute to the field of candlestick data research.