Clickstream String Processing with Suffix Arrays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing clickstream data, such as Markovian processes, are inefficient and limited in identifying complex patterns across multiple pages, leading to delayed insights and less effective website improvements.
Innovation Solution
The application of string processing techniques, including assigning unique symbols to website pages, generating a master string, suffix array, and longest common prefix array, to identify the most common n-step subsequences of pages visited by users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional Markovian processes are used to analyze clickstream data, then the analysis can identify basic user behavior patterns, but the analysis time and processing resources increase substantially as the size of the website and number of users increases
Solution Approach 1:
The patent segments the clickstream data analysis by converting it into a string processing problem. Each user session is represented as a string of page symbols, and the master string concatenates all session strings with unique delimiters. This segmentation allows the use of efficient suffix array algorithms to identify patterns without the computational overhead of traditional Markovian processes, directly resolving the contradiction between pattern identification accuracy and analysis time.
Solution Approach 2:
The patent replaces the mechanical computational approach of traditional Markovian analysis with a more efficient string processing mechanism. By substituting the Markov chain computational model with suffix array and LCP array data structures, the system achieves the same pattern identification function with significantly reduced time complexity, addressing the time loss issue while maintaining measurement precision.
2Measurement precision
If traditional Markovian processes are used to analyze clickstream data, then basic behavior patterns can be identified, but processing resources and storage requirements become excessive for large-scale websites
Solution Approach 1:
The patent segments the complex clickstream analysis into discrete string operations. By representing each page as a symbol and each user session as a string, the system breaks down the analysis into manageable units that can be processed efficiently using suffix arrays. This segmentation reduces the memory footprint and processing requirements compared to maintaining complex Markov transition matrices for large-scale websites.
Solution Approach 2:
The patent changes the parameter representation from continuous probability distributions in Markov models to discrete string symbols. This parameter transformation allows the use of compact suffix array data structures that require less storage and processing resources while maintaining the ability to detect behavior patterns accurately.
3Loss of information
If more comprehensive user behavior data is collected to improve insight quality, then the amount of data to be analyzed increases substantially, making efficient analysis difficult
Solution Approach 1:
The patent applies preliminary action by pre-processing the clickstream data into a standardized string format during data collection. By converting page sequences into symbolic strings and building suffix arrays in advance, the system prepares the data structure to enable rapid pattern searching and analysis. This preliminary transformation maintains comprehensive user behavior information while significantly improving subsequent analysis efficiency.
Solution Approach 2:
The patent creates a symbolic copy of the clickstream data where actual page URLs are replaced with compact page symbols. This copying approach preserves all user behavior patterns and insights while reducing the data size and complexity, enabling efficient analysis of large-scale comprehensive behavior data without sacrificing information quality.
Data Source
AI summary
A method includes assigning unique symbols to pages of a website, respectively. The method includes obtaining page symbol sequences of browsing sessions, respectively. Each browsing session corresponds to a visitor of the website. For each browsing session, the page symbol sequence of the browsing session is a sequence of symbols that corresponds, respectively, to a sequence of pages of the website visited during the browsing session by the corresponding visitor. The method includes generating a master string including the page symbol sequences, generating a suffix array corresponding to the master string, and generating a longest common prefix (LCP) array corresponding to the suffix array. The method includes, based on the suffix array and LCP array, determining one or more most common n-step subsequences of pages (n is an integer greater than 1).