Tag: on-chain modeling

Annotation: The relevance of the study is determined by the need to create tools for objective identification of digital asset market phases based on the behavioral characteristics of network participants, which are recorded not by price, but by on-chain metrics. In the current environment of growing cryptocurrency market volatility and the insufficient effectiveness of classic technical analysis indicators, there is an increasing demand for algorithms capable of detecting latent market states, in particular the accumulation phase, even before price changes occur. The purpose of this article is to create an algorithmic model for detecting the accumulation phase in the cryptocurrency market by applying cluster analysis to a set of on-chain indicators, which allows for more accurate and timely assessment of market dynamics for use in investment decision support systems. The research methodology is based on the use of blockchain network behavioral metrics (Realized Cap HODL Waves, Dormancy, SOPR, Exchange Outflow Volume, Address Balance Distribution) and unsupervised cluster learning algorithms. HDBSCAN was used as the main clustering method, allowing for the adaptive identification of market phases without fixing their number. A procedural diagram was constructed covering the stages of data collection, processing, normalization, cluster distribution, and interpretation of results. The research results reflect the construction of an effective algorithm capable of grouping time intervals by similarity of on-chain behavior and identifying accumulation phases taking into account the time dynamics of metrics. A classification architecture has been implemented that does not depend on price data and can function in real time as part of cryptanalysis systems. The conclusions prove the effectiveness of the proposed model for detecting accumulation phases based on the dynamics of on-chain data structures. It has been established that the algorithm is capable of ensuring early detection of changes in market behavior and reducing dependence on traditional speculative indicators. Prospects for further research are related to the integration of time dependencies into the cluster model, the extension of the system to other digital assets, the introduction of aggregated accumulation indices and self-learning components, taking into account macroeconomic and off-network influences.