Unsupervised Patron Clickstream Clustering With UMAP And DBSCAN
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing patron study and survey methods for analyzing human interactions are limited in scope, costly, and rely on known hypotheses, making them inefficient and prone to skewed results due to participant selection biases.
Innovation Solution
An apparatus and method utilizing unsupervised machine learning algorithms, including UMAP, DBSCAN, and KNN, to analyze patron clickstream data for clustering without prior hypotheses, transforming data into a probability matrix, reducing dimensionality, and visually illustrating patron segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional patron surveys and studies are used to analyze human interactions, then results can be obtained with known hypotheses, but the method is limited in scope, costly, and prone to participant selection biases
Solution Approach 1:
The system performs unsupervised clustering autonomously without requiring human-designed hypotheses or survey questions. The algorithm automatically discovers patterns in clickstream data, allowing the data itself to serve the analysis rather than requiring external guidance from researchers.
Solution Approach 2:
The patent replaces the mechanical survey process (designing questions, recruiting participants, collecting responses) with an automated computational system that processes clickstream data through machine learning algorithms, eliminating human involvement in data collection and hypothesis formulation.
2Loss of information
If patron surveys are conducted to gather interaction data, then behavioral information can be collected, but the cost and time required for conducting and analyzing results increases
Solution Approach 1:
The system collects and stores clickstream data continuously in the background before analysis is needed. This preliminary data collection eliminates the time required for survey administration and allows for immediate or scheduled analysis without waiting for data gathering.
Solution Approach 2:
Instead of collecting raw survey responses that require extensive processing, the system uses pre-processed clickstream metrics (page views, time on page, navigation patterns) that are already in an analyzable format, effectively using a copy of the behavioral information in a more efficient form.
3Loss of information
If surveys rely on known questions and hypotheses, then focused insights can be obtained, but unknown behaviors cannot be anticipated and the method is limited in scope
Solution Approach 1:
Instead of starting with hypotheses and testing them against data, the system inverts the approach by letting the data generate its own patterns and categories through unsupervised clustering. The analysis emerges from the data rather than being imposed upon it.
Solution Approach 2:
The system transforms high-dimensional clickstream data into a probability matrix and then applies dimensionality reduction to reveal underlying patterns. By changing the representation parameters of the data, the system can discover behaviors that would not be apparent in the original format.
4Loss of information
If high dimensional patron clickstream data is analyzed directly, then comprehensive behavior patterns can be captured, but the data complexity and computational requirements increase
Solution Approach 1:
The system transforms high-dimensional clickstream data into a probability matrix and then reduces it to two-dimensional visualizations. This dimensional transformation preserves the essential behavioral patterns while making the data computationally manageable and visually interpretable.
Solution Approach 2:
The system extracts the essential clustering structure from the high-dimensional data by identifying cluster centers and visualizing only the relevant features. This extraction separates the signal (behavioral patterns) from the noise (extraneous dimensions).
Data Source
AI summary
Groups of patrons may be discovered by measuring website and mobile site patron clickstream data in a mathematical and unsupervised way over a predetermined time and by graphically clustering the patron clickstream data using non-linear dimensionality reduction in the form of a Uniform Manifold Approximation and Projection algorithm (UMAP). The data from the UMAP may then be feed into a Density Based Spatial Clustering of Applications with Noise algorithm (DBSCAN) in order to identify a center of each cluster. Next, using the data from the UMAP and the center of each cluster from the DBSCAN, a K-Nearest Neighbor algorithm (KNN) may be applied to identify data points closest to the center of each cluster and to shade each of the data points to graphically identify each cluster of the plurality of clusters. Next, illustrate a graph on the display representative of the data points shaded following application of the KNN.