Decision Tree Event Prediction Using Clickstream Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to effectively predict the outcomes of user-requested events during software application sessions, leading to potential degradation in user experience due to event failures, such as network connection issues.
Innovation Solution
A method and system that extract historical user events from clickstream data, generate features, and train a decision tree model to predict event outcomes by calculating feature relevance metrics and traversing a path through the tree based on new user events, using a combination of user and machine features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a decision tree model is trained using historical clickstream data and user features, then prediction accuracy is improved (93% accuracy), but device complexity and computational resources increase
Solution Approach 1:
The system performs preliminary actions by extracting and processing historical clickstream data, generating user features, and training the decision tree model in advance. This preparation work is done offline before actual event prediction is needed, allowing the model to be ready for high-accuracy predictions without adding real-time computational complexity.
Solution Approach 2:
The system creates a simplified copy of user behavior patterns through the decision tree model. Instead of analyzing raw clickstream data in real-time, the system uses the trained model which captures essential patterns, enabling accurate predictions with reduced computational requirements during actual events.
2Reliability
If feature relevance metrics are calculated and multiple user features are used for prediction, then prediction reliability is improved, but loss of time during data processing increases
Solution Approach 1:
Feature relevance metrics are calculated in advance during the model training phase using historical data. This preliminary calculation identifies the most important features before actual prediction is needed, so during real-time events, only these pre-identified features need to be processed, reducing time loss while maintaining reliability.
Solution Approach 2:
The system extracts and focuses only on the most relevant features identified through feature relevance metric calculation. By taking out and concentrating on these key features rather than processing all possible features, the system maintains high prediction reliability while reducing data processing time.
Data Source
AI summary
A method may include extracting, from historical clickstream data, historical user events, generating historical user features from the historical user events, generating, using the historical user features, a training data set including feature vectors each labeled with an outcome of a historical requested event, adding, to a decision tree: internal nodes corresponding to the historical user features, branches corresponding to the values of the historical user features, and leaf nodes each corresponding to an outcome of the historical requested event. The method may further include extracting, from new clickstream data, new user events, generating new user features from the new user events, detecting a request to cause a new requested event, and predicting an outcome of the new requested event by traversing a path including: a series of branches each corresponding to a value of one of the new user features, and a leaf node corresponding to the outcome.