Automatic log file analysis using Machine Learning techniques for better understanding user behaviour

Question

I have to automatically analyse json log files using Machine Learning techniques in order to better understand the user behavior and recognise frequently performed actions to improve the usability of a application. The log file contains entries like Filesystem Events, Mouse Events, Keyboard Event, Application Events (launched, terminated, active, deactivated, ...), ... which were logged during the usage of a specific application.

The Logged entries look like this:

       {
        "timestamp" : "20150427T100724.571637",
        "entry" : 
        {
            "eventType" : 1,
            "eventTypeString" : "created",
            "nodePath" : "/Applications/Firefox.app",
            "nodeType" : 3,
            "nodeTypeString" : "directory"
        },
        "entryType" : 4
    },

    {
        "timestamp" : "20150427T100729.072595",
        "entry" : 
        {
            "eventType" : 2,
            "eventTypeString" : "clickRight",
            "mouseType" : 1,
            "mouseTypeString" : "local",
            "positionX" : 2517,
            "positionY" : 166,
            "scrollWheel" : 0.0
        },
        "entryType" : 1
    },

    {
        "timestamp" : "20150427T100730.270000",
        "entry" : 
        {
            "eventType" : 2,
            "eventTypeString" : "unmounted",
            "volumeName" : "Firefox",
            "volumePath" : "/Volumes/Firefox"
        },
        "entryType" : 9
    },

    {
        "timestamp" : "20150427T100730.281202",
        "entry" : 
        {
            "eventType" : 2,
            "eventTypeString" : "removed",
            "nodePath" : "/Volumes/Firefox",
            "nodeType" : 3,
            "nodeTypeString" : "directory"
        },
        "entryType" : 4
    },

    {
        "timestamp" : "20150427T100732.668000",
        "entry" : 
        {
            "eventType" : 4,
            "eventTypeString" : "activated",
            "launchDate" : "20150427T100732.668000",
            "ownsMenuBar" : true,
            "processId" : 7011,
            "processName" : "TeamViewer"
        },
        "entryType" : 5
    },

I already got into the topic of machine learning but i have no idea which specific technique or even sub area of the field could help me solving this. Maybe someone has a hint or idea where to find the right information or concept for a solution?

score 0 · Answer 1 · answered May 09 '15 at 18:17

Okay the question is not totally clear to me but whatever i have understood based on that these are my opinions.

Classification , not sure how it can be applied here.
Clustering same as above ,
Lot of scope for probability specifically Probabilistic graphical models , ex in the file system if he opened c:// then songs folder most probably his next folder choice can be Rihanna folder or Emineem, these stuffs to an extent can be modeled by graphical models. You can make your custom models. it is a hugh topic , check out this

https://class.coursera.org/pgm/lecture

Automatic log file analysis using Machine Learning techniques for better understanding user behaviour

1 Answers1