In the finance sector, time series forecasting has been widely used in situations such as stock prices forecasting and equity market forecasting. In recent years, techniques of computer vision include become increasingly popular in the monetary assumption of periodicity. A hot subject is how to mark information from macroeconomic variables to assess the accuracy rate of computer vision techniques and, as a result, final investment returns. Existing information from macroeconomic variables labelling methods primarily mark data and comparing current data to data from a limited period in the future. Information from a macroeconomic variable, on the other hand, are often quasi, with apparent unpredictability in the brief period. As a result, such tagging methods struggled to identify the time series analysis information's persistent trend functionality, resulting in a misalignment among their pigeonhole results and actual market inclination. A new pigeonhole approach called 'continuous trend labeling' is proposed in this paper to fix the issue. This article presents a novel feature preprocessing tack that avoids taking a gander distortion that plagues continuous data consistency and preprocessing approaches. The principle of constant trend marking and an advanced tagging framework for removing consistent trend characteristics from time-series data were then presented, along with a detailed logical explanation. Experiments on the ensembled model yielded a scoring accuracy of 52 percent, which was close to the Cagliari teams. The results were determined by combining all of the networks' ratings. Even though it appears to be a low percentage, being right on a trade position more than 50% of the time is considered good, particularly when looking at the data in isolation. Our labelling system is significantly superior to classifiers and other identification evaluation measures; province tagging approaches are used. According to the observations of the study, word embeddings like LSTM and GRU are suitable to finance related time series data prediction.
|Idioma original||Inglés estadounidense|
|Título de la publicación alojada||Proceedings - 2nd International Conference on Smart Electronics and Communication, ICOSEC 2021|
|Editorial||Institute of Electrical and Electronics Engineers Inc.|
|ISBN (versión digital)||9781665433686|
|Estado||Indizado - 2021|
|Evento||2nd International Conference on Smart Electronics and Communication, ICOSEC 2021 - Trichy, India|
Duración: 7 set. 2021 → 9 set. 2021
Serie de la publicación
|Nombre||Proceedings - 2nd International Conference on Smart Electronics and Communication, ICOSEC 2021|
|Conferencia||2nd International Conference on Smart Electronics and Communication, ICOSEC 2021|
|Período||7/09/21 → 9/09/21|
Nota bibliográficaPublisher Copyright:
© 2021 IEEE.