Graph Definitions

1
import kumoai.rfm as rfm
2
3
# Method 1: Utility function (recommended for most cases)
4
# Automatically creates tables from data frames, infers metadata, and finds links
5
graph = rfm.Graph.from_data({
6
    'users': df_users,
7
    'products': df_products,
8
    'transactions': df_transactions
9
})
10
11
# Method 2: Manual construction from pre-configured table objects
12
tables = [users_table, products_table, transactions_table]
13
graph = rfm.Graph(tables=tables)
14
graph.infer_links()  # or define links manually

1
# For example, these column patterns create automatic links:
2
# transactions.user_id -> users.user_id (or users.id)
3
# orders.product_id -> products.product_id (or products.id)
4
# reviews.customer_id -> customers.customer_id (or customers.id)
5
6
# View inferred edges
7
for edge in graph.edges:
8
    print(f"{edge.src_table}.{edge.fkey} -> {edge.dst_table}")

1
# Add specific edge
2
graph.link(src_table="transactions", fkey="user_id", dst_table="users")
3
4
# Remove edge
5
graph.unlink(src_table="transactions", fkey="user_id", dst_table="users")

1
# Original table mixing transaction, bank, and user data
2
mixed_data = pd.DataFrame({
3
    'transaction_id': [1, 2, 3],
4
    'bank_id': [101, 102, 101],
5
    'user_id': [201, 202, 203],
6
    'transaction_amount': [100.0, 250.0, 75.0],
7
    'transaction_type': ['deposit', 'withdrawal', 'transfer'],
8
    'bank_name': ['Chase', 'Wells Fargo', 'Chase'],
9
    'bank_routing': ['123456', '789012', '123456'],
10
    'user_name': ['Alice', 'Bob', 'Charlie'],
11
    'user_email': ['alice@email.com', 'bob@email.com', 'charlie@email.com']
12
})
13
14
# Split into three entity-focused tables
15
16
# 1. Transactions table (transaction-specific data)
17
transactions = mixed_data[['transaction_id', 'bank_id', 'user_id', 'transaction_amount', 'transaction_type']].copy()
18
19
# 2. Banks table (bank-specific data)
20
banks = mixed_data[['bank_id', 'bank_name', 'bank_routing']].drop_duplicates()
21
22
# 3. Users table (user-specific data)
23
users = mixed_data[['user_id', 'user_name', 'user_email']].drop_duplicates()
24
25
# Create graph with proper entity relationships
26
graph = rfm.Graph.from_data({
27
    'transactions': transactions,
28
    'banks': banks,
29
    'users': users
30
})
31
# Result: transactions.bank_id -> banks.bank_id and transactions.user_id -> users.user_id

1
# Problem: Table with many-to-many data stored as lists/comma-separated values
2
user_skills_combined = pd.DataFrame({
3
    'user_id': [1, 2, 3],
4
    'user_name': ['Alice', 'Bob', 'Charlie'],
5
    'skills': [['Python', 'SQL'], ['SQL', 'Machine Learning'], ['Python', 'Machine Learning']],
6
    'proficiency_levels': [['expert', 'beginner'], ['intermediate', 'advanced'], ['expert', 'expert']]
7
})
8
9
# This structure cannot create proper foreign key relationships in KumoRFM
10
11
# Solution: Normalize into three tables with junction table
12
13
# 1. Users table (entity table)
14
users = user_skills_combined[['user_id', 'user_name']].copy()
15
16
# 2. Skills table (entity table)
17
all_skills = []
18
for skill_list in user_skills_combined['skills']:
19
    all_skills.extend(skill_list)
20
unique_skills = list(set(all_skills))
21
22
skills = pd.DataFrame({
23
    'skill_id': range(1, len(unique_skills) + 1),
24
    'skill_name': unique_skills
25
})
26
27
# 3. Junction table (breaks many-to-many into two one-to-many)
28
user_skills_records = []
29
for _, row in user_skills_combined.iterrows():
30
    for skill, proficiency in zip(row['skills'], row['proficiency_levels']):
31
        skill_id = skills[skills['skill_name'] == skill]['skill_id'].iloc[0]
32
        user_skills_records.append({
33
            'user_skill_id': len(user_skills_records) + 1,
34
            'user_id': row['user_id'],
35
            'skill_id': skill_id,
36
            'proficiency_level': proficiency
37
        })
38
39
user_skills = pd.DataFrame(user_skills_records)
40
41
# Create graph with proper one-to-many relationships
42
graph = rfm.Graph.from_data({
43
    'users': users,
44
    'skills': skills,
45
    'user_skills': user_skills
46
})
47
# Result: user_skills.user_id -> users.user_id and user_skills.skill_id -> skills.skill_id

1
graph.visualize()

1
graph.validate()