Do you have a massive amount of data and are wondering how to make sense of it? Are you looking for ways to break down that data into a logical structure?
What if we told you there’s a machine learning algorithm that can automate this process for you?
Exciting, right? Let’s first explore how to manually create a KPI Tree from a given dataset.
Imagine having a dataset of customers who visit an e-commerce website frequently. By analyzing their web-navigation attributes like device category, browser, visit source, and geo-network region, we aim to predict whether they will convert or not.
Let’s create a sample dataset of 100 records:
import pandas as pd
import random
# Sample data for each column
device_categories = ['Desktop', 'Android Mobile', 'Apple Mobile', 'Tablet', 'Laptop']
browsers = ['Chrome', 'Opera', 'Safari', 'Firefox', 'Edge']
visit_sources = ['Paid Social', 'Referral', 'Organic Search', 'Direct', 'Email Campaign']
geonetwork_regions = ['Mumbai', 'Delhi', 'Bangalore', 'Hyderabad', 'Chennai', 'Kolkata', 'Pune', 'Ahmedabad', 'Surat', 'Jaipur']
conversion_status = [0, 1]# Generate random data
data = {
'device_category': [random.choice(device_categories) for _ in range(100)],
'browser': [random.choice(browsers) for _ in range(100)],
'visit_source': [random.choice(visit_sources) for _ in range(100)],
'geonetwork_region': [random.choice(geonetwork_regions) for _ in range(100)],
'will_convert': [random.choice(conversion_status) for _ in range(100)]
}
# Create DataFrame
df = pd.DataFrame(data)
# Display the records
df
Here’s a snapshot of the data:
As we aim to predict whether customers will convert or not, we dive into a classification/prediction problem. This is where the concept of a KPI Tree comes in, offering a more structured approach than a traditional Decision Tree Classifier.
A KPI Tree breaks down data into a logical structure, allowing flexibility in node hierarchy based on business requirements. While it differs from a conventional Decision Tree, which predicts class labels at leaf nodes based on metrics like entropy and information gain, the KPI Tree offers a unique perspective.
In essence, a KPI Tree dissects your data like a decision tree but with random node order and hierarchy, tailored to your analysis needs.
For example, in an e-commerce scenario, you might prioritize geo-network region, followed by device category, browser, and visit source in your KPI Tree structure.
This approach can unveil insights in a hierarchy, such as conversion rates across regions, devices, browsers, and visit sources.
Evolving from a basic Decision Tree, a KPI Tree offers a dynamic narrative of your data’s intricate relationships, guiding you through diverse paths and combinations.