Challenge 1: Credit Scoring

The training dataset includes all addresses that had undertaken at least one borrow transaction on Aave v2 Ethereum or Compound v2 Ethereum any time between 7 May 2019 and 31 August 2023, inclusive (called the observation window).

Data Structure & Shape

There are almost 0.5 million observations with each representing a single borrow event. Therefore, all feature values are calculated as at the timestamp of a borrow event and represent the cumulative positions just before the borrow event's timestamp.

Given that the data is at the borrow transaction level, an address with 10 borrow events during the observation window will appear 10 times in the dataset with different feature values across the 10 borrow timestamps.

As an example, if address 0xa initiated the following transactions on Aave v2 Ethereum:

• At block_0: deposit 100 ETH

• At block_1: borrow 25 ETH

• At block_3: repay 10 ETH

• At block_4: borrow 10 ETH

then 0xa will appear twice in the dataset as follows:

• At block_1: all feature values will be calculated using all Ethereum and wallet transactions up to block_0, i.e., excluding the borrow event and other transactions at block_1

• At block_4: all feature values will be calculated using all Ethereum and wallet transactions up to block_3, i.e., excluding the borrow event and other transactions at block_4

The dataset has 79 features in total, including the target label, transaction block number, transaction timestamp, and ETH address of the borrower.

Features

Other than counts related features, all feature values are denominated in ETH. All price conversions of the underlying tokens to ETH were done with the applicable rates as obtained from Transpose, Coinpaprika, CoinGecko, and Compound and Aave's price feeds at the respective event timestamps.

Recommended 3rd party data sources

Some recommended data sources specific to Challenge 1 include:

• Transpose

• Etherscan

• Subgraphs

  • Aave v2 Ethereum

  • Compound v2 Ethereum

• Uniswap

However, it is imperative that any data other than those provided by the Spectral SDK should be readily available in a timely manner during Consumption Window to ensure timely real-time predictions.