However, these tools are constrained in respect to acquiring, managing and the analysis of web-based data as a multi-V model to express data sets that are characterized by a massive volume, a variety of features and high velocity. Such analytical features refers to big data analytics techniques that have the ability to process data with an immense volume (from gigabytes to terabytes), variety (from semi-structured to unstructured) and velocity (from batch to streaming) via unique data storage management, analysis and visualization technologies. Thus, the differences in analytical capability between big data architecture and traditional web analytics architecture, are that the former has a unique ability to analyze unstructured textual data, to parallel process large data volumes, and to parse data in real time or near real time.
Moreover, the current development of machine learning algorithms, NLP-based text mining, data mining and graph mining also suggest a new direction of web analytics which more is geared toward predictive and prescriptive analytics as opposed to descriptive analytics. A major emerging component in this new era of web analytics applications is the development of cloud computing platforms and services delivered as services (i.e., SaaS) over the Internet. Cloud services for web analytics have become crucial given that communication costs often dominate computation costs, and thus it is essential to move the analytics capabilities ‘closer’ to the data where it is collected and stored.
Adding additional insights into the business impact of state-of-the-art web analytic and the evolution to digital intelligence propel a major shift from website-centricity tactics toward extended practices i.e. – to a more strategic (Big) Data analytics framework.
This framework aims to provide a complete solution in this space of emerging digital intelligence, in diverse analytical applications such as social network analysis (e.g., SNA, organizational entities analysis), text analytics (e.g., information extraction, sentiment analysis), advanced visualization techniques (e.g., co-occurrences networks, words cloud), clustering algorithms, anomaly detection mechanisms , trend analysis, pattern discovery and predictive modeling methods (e.g. decision trees, random forest, logistic regression, SVM).
This approach to web analytics brings a set of extended requirements, augmented metrics, and optimization practices to the current base of web analytics and will be covered more deeply in the forthcoming series of posts on this topic.
Dr. Elan Sasson. Intlock LTD