Tools for Social Network Analysis with the Focus on Behavioural Patterns

IR_Moodle

Data preprocessing of the log from LMS Moodle.

Input:
  • Log file from multi-agent system
  • courses.csv - file of course names, which will be filtered
  • eventPrefixes.csv - file of activities, which will be filtered
Output:
  • graph.gdf - file of sequence graph, in which all the extracted sequences are represented as graph nodes
  • profile.Base.csv - file of Base agent profiles. The file contains the agents and their sequences of events, extracted from the log file.

IR_Agents

Data preprocessing of the log from multi-agent system.

Input:
  • Log file from multi-agent system
Output:
  • graph.gdf - file of sequence graph, in which all the extracted sequences are represented as graph nodes
  • profileBase.csv - file of Base student profiles. The file contains the students and their sequences of events, extracted from the log file.

IR_Web

Data preprocessing of the log from Apache web server.

Input:
  • Apache web log server
Output:
  • graph.gdf - sequence graph, in which all the extracted sequences are represented as graph nodes
  • profile.Base.csv - file of base user profiles. The file contains the users and their sequences of events, extracted from the log file.

IR_GraphReweighting

Weighting of tie strength between the sequences in the basic sequence graph and sequence clustering. The process consists of three steps:

  1. Counting the weights between the sequences. The output is influenced by two parameters: (1) the method used for determination of sequence similarity (weighting), which can be selected from methods LCS, LCSS, T-WLCS, Levenshtein distance, or cosine measure. (2) the minimal weight between the sequences.
  2. Sequence clustering using graph partitioning by Left-Right Oscillate algorithm based on spectral clustering method or using hierarchical agglomerative clustering.
  3. A more precise division of the obtained sequence clusters and finding representatives for these clusters.
Input:
  • graph.gdf - file of sequence graph, in which all the extracted sequences are represented as graph nodes
Output:
  • graph@last@.gdf.cmp - file of sequence clusters for selected input weight threshold
  • file of sequence clusters and their representatives
  • .gdf file - the graph with colored new sequence clusters and the relations between them
  • profileReduced.csv - file of reduced user profiles. The file contains the students and their sequences of events, extracted from the log file.

IR_UsersGraph

Creation of network user groups with similar behaviour. The similar behaviour can be determined by base or reduced user profiles. Two approaches can be used for construction of users' network. (1) The first method proceeds from graph of users (a .gdf file). The groups of users with similar behaviour (similar user profiles) are found using spectral clustering by Fiedler vector using the Left-Right Oscillate algorithm. (2) The second method uses Self organizing maps (SOM) for the construction of users’ network containing groups of users with similar behaviour.

IR_Parser

Tool for graph conversion. Supported are formats like: .gdf, .csv, .ijv.

Site Map :: Printable View :: © 2008 - 2017 VSB-TU Ostrava :: Powered by mojoPortal :: HTML 5 :: CSS :: Design by Graformix