Clickstream Sequence Visualization for E-commerce Behavior Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing customer behavior in e-commerce, such as gathering clickstream data, are inefficient and rely heavily on human expertise due to the presence of useless data and the complexity of identifying useful behavior models.
Innovation Solution
A data visualization method that captures clickstream data, compares segment patterns using a sliding window method to determine similarity, and visualizes the most similar sequences in a 2D space, where the order and event status of click data are mapped to dimensions, allowing for automated identification of patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complete clickstream data is captured and analyzed, then comprehensive customer behavior information is obtained, but data complexity and volume increase significantly making analysis inefficient
Solution Approach 1:
The patent segments the complete clickstream data into multiple sequence segments based on behavioral patterns and event types. By dividing the large volume of clickstream data into smaller, manageable segments, the system can analyze each segment independently to identify useful behavior models without being overwhelmed by the entire dataset at once.
Solution Approach 2:
The patent extracts and filters out useful behavior models from the complete clickstream data by comparing sequence segments against predefined patterns. This extraction process removes irrelevant and useless data sections, retaining only the meaningful behavioral patterns that contribute to accurate customer behavior analysis.
2Measurement precision
If specialists manually check clickstream contents to find useful behavior models, then accurate behavior patterns can be identified, but the process has low efficiency and high dependency on human expertise
Solution Approach 1:
The patent implements an automated system that performs behavior model identification without requiring specialist intervention. The system automatically captures clickstream data, segments it into sequence segments, compares segments against patterns, and identifies useful behavior models through computational processes, making the analysis self-service and independent of human expertise.
Solution Approach 2:
The patent replaces the manual mechanical process of specialists checking clickstream contents with an automated computational system. The system uses algorithmic comparisons between sequence segments and patterns to automatically identify behavior models, substituting human cognitive work with mechanical computation that is faster and more scalable.
3Productivity
If conventional statistical methods are used to analyze clickstream data, then basic metrics can be computed, but accurate customer behavior survey is not achieved
Solution Approach 1:
The patent transitions from static statistical methods to a dynamic sequence-based analysis approach. By treating clickstream data as ordered sequences of events and comparing them against pattern sequences, the system captures the temporal dynamics and progression of customer behaviors, providing more accurate insights than static statistical aggregations.
Solution Approach 2:
The patent changes the analytical parameters from conventional statistical metrics to sequence pattern matching criteria. Instead of computing traditional statistics like averages and distributions, the system evaluates the similarity between sequence segments and behavioral patterns, fundamentally changing the measurement parameters to better capture customer behavior nuances.
Data Source
AI summary
A data visualization method and a data visualization device. The data visualization method includes the following steps: capturing a clickstream that includes a plurality of click data; generating a similarity value for each of the plurality of click data by comparing a first sequential segment of each of the plurality of click data with a segment pattern; capturing the click data having the maximum similarity among the plurality of click data and capturing a second sequence segment of each of the click data having the maximum similarity; visualizing the second sequential segments in a 2D space to present the visualized sequence data of each of the second sequence segments, and setting a position of the click data, having the maximum similarity, in the visualized sequence data to be at a datum point in a first dimension of the 2D space.