การทำนายแนวโน้มของราคาบิทคอยน์
- ชื่อนักเรียนผู้จัดทำโครงงานวิทยาศาสตร์
กษิดิ์เดช อิ้วศรีสกุล
- อาจารย์ที่ปรึกษาโครงงานวิทยาศาสตร์
กอบชัย ดวงรัตนเลิศ
- โรงเรียนที่กำกับดูแลโครงงานวิทยาศาสตร์
- ปีที่จัดทำโครงงานวิทยาศาสตร์
บทคัดย่อโครงงานวิทยาศาสตร์
Introduction:
In the fast-paced realm of cryptocurrency, the fluctuations and turbulence of Bitcoin have become increasingly pronounced. In this landscape, the need for dependable tools for analyzing prices is paramount for investors. However, the predicament lies in my struggles to acclimate to unfamiliar future data, shaped by evolving contexts. While the core principles remain steadfast, I present an innovative domain adaptation approach tailored for the prediction of Bitcoin trends.
Method:
My method hinges on a systematic process of curating minute-level Bitcoin data spanning from 2014 to 2023. To enhance the data's utility, I normalize it using logarithmic transformations and subsequently classify it into three distinct trends: 'up,' 'down,' and 'neutral.' This pioneering solution is realized through the careful partitioning of data into training/testing source sets and training/testing target sets. Central to my approach is a tripartite model architecture, comprising an encoder, decoder, and self-attention mechanisms.
Conclusion:
The efficacy of my approach is corroborated by the validation performed on a reserved holdout test set, revealing an impressive classification accuracy that surpasses 60% for both source and target sets. The credibility of my methodology is further substantiated through rigorous backtesting, underscoring its potential for profitability. Through these cumulative insights, my method emerges as a promising step forward in achieving resilient and adaptive cryptocurrency trend prediction.