Algorithm 1: Dataset splitting based on geographic region hashing (hash bucketing)

Input: region: string   // Geographic region name

Output: subset: {train, val, test}

1: // Step 1: Generate normalized hash value

2: hash_str ← MD5(region)                       // Compute MD5 hash of region

3: hash_hex ← Substring(hash_str, 0, 8)   // Extract first 8 hexadecimal characters

4: hash_int ← HexToInt(hash_hex)           // Convert hexadecimal substring to integer

5: hash_mod ← hash_int MOD 10,000     // Limit value range via modulo operation

6: hash_ratio ← hash_mod/10,000.0       // Normalize to [0, 1)

7: // Step 2: Assign subset based on ratio

8: if hash_ratio < 0.8 then

9:          subset ← “train”          // 80% for training set

10: else if hash_ratio < 0.9 then

11:          subset ← “val”          // 10% for validation set

12: else

13:          subset ← “test”          // 10% for testing set

14: end if