Relational Neighbor Classifier

In this exercise, you will apply a simple network based classifier called the relational neighbor classifier. It uses the class labels of neighboring nodes to compute a churn probability for each node in the network.
For example, in the network below where red nodes denote churners and white nodes denote non-churners, the churn probability of the blue node is 0.4.

Relational neighbor classifier

You are given two vectors: ChurnNeighbors and NonChurnNeighbors with each customer's number of neighbors that have churned and not churned, respectively.

Diese Übung ist Teil des Kurses

Predictive Analytics using Networked Data in R

Kurs anzeigen

Anleitung zur Übung

Compute the churn probability of each customer, churnProb, using the relational neighbor classifier.
Use which() to find the customers with the highest probability of churning. Call this vector mostLikelyChurners.
Use mostLikelyChurners to find the IDs of the customers with the highest churn probability.

Interaktive Übung

Versuche dich an dieser Übung, indem du diesen Beispielcode vervollständigst.

# Compute the churn probabilities
churnProb <- ___ / (ChurnNeighbors + ___)

# Find who is most likely to churn
mostLikelyChurners <- which(churnProb == ___(churnProb))

# Extract the IDs of the most likely churners
customers$id[___]

Code bearbeiten und ausführen