Detect Areas of Upward and Downward Fluctuations in Bitcoin Prices Using Patterned Datasets

Abstract

Many researchers have proposed various theories and techniques to estimate the value of Bitcoin. However, there is still a great opportunity to suggest a new approach to Bitcoin forecasting. This research will examine the concept of forming a patterned dataset to prove the hypothesis regarding the start of a bitcoin price fluctuation until the end of the fluctuation using patterned dataset pairs in conditions of price decline (moon) and in conditions of price decline (crash). Patterned datasets were initially built from data on the movement of all cryptocurrency assets on a digital cryptocurrency trading market. The Patterned Dataset is programmed based on the response of all crypto assets to changes in the price of bitcoin over time. Patterned datasets are better able to describe the conditions of the digital cryptocurrency market, along with patterns of potential increases and decreases in Bitcoin prices in particular and cryptocurrencies in general. To deepen the observations, resampling of the patterned dataset was carried out with a 4-hour timeframe using the mean, median, min, max, and sum filter functions on pairs of crash and moon conditions per month. The results show that filter functions other than sum, namely mean, median, min, and max, provide markings of fluctuation area patterns that are easier to recognize. Two hypotheses related to fluctuations were also successfully proven in this research.