Treasury Asset Monitoring & Evaluation
Asset monitoring
The significance of the DAO treasury cannot be underestimated; it plays a central role in the DAO ecosystem. We employ Alchemy and official APIs to monitor on-chain data from numerous mainstream blockchains. Furthermore, we analyze data from prevalent protocols in the market. This approach ensures precise coverage of DAO's asset data across various chains and protocols, including detailed information on the inflow and outflow of funds.
Asset evaluation
To conduct a comprehensive analysis of DAO assets, we have initially gathered data on these assets. For fungible token (FT) assets, we have opted to utilize prices from CoinGecko as the evaluation basis. As for non-fungible token (NFT) assets, we have developed an independent evaluation framework to ensure effective assessment of this asset class.
The NFT asset evaluation framework
To comprehensively evaluate NFT assets, obtaining transaction data is crucial. We achieved this by utilizing official APIs and parsing on-chain data from issuance and transaction contracts to collect data from major NFT issuance and trading platforms.
Within our evaluation framework, we categorize NFT assets into pfp type, 1155 type, collection type, art type, openstore type, openstore1155 type, and ens type based on their characteristics. To determine assessment prices, we employ a unified core K_Means clustering algorithm.
For pfp-type assets, which include feature data, we conduct separate clustering based on different features during evaluation to obtain accurate assessment prices. As for 1155-type, collection-type, and art-type assets, we utilize the most recent transaction prices for each NFT. Through clustering in both price and time dimensions, we then take the median price of the "recent time class" to represent the assessment price for a particular category of assets.
K_Means clustering algorithm
Clustering is a technique in unsupervised learning where, in the absence of any prior knowledge of sample labels, samples are partitioned into several categories based on inherent relationships within the data. This process aims to maximize the similarity among samples within the same category while minimizing the similarity between samples from different categories (i.e., increasing intra-cluster cohesion and reducing inter-cluster separation).
Clustering is a form of unsupervised learning, and K-Means clustering stands out as one of the fundamental and commonly used clustering algorithms. The underlying concept involves iteratively searching for a partitioning scheme of the data into K clusters, where the clustering result corresponds to the minimum value of the associated loss function.
Last updated