Abstract
The present work involves the development of a learning model for the analysis of data from telecommunication and social networks for devices that make voice calls and data sessions, using Business Intelligence concepts and tools. The paper has a multi/interdisciplinary character, calling for e-learning, telecommunication, database, programming, IT applications and social/business networks.
Business Intelligence (BI) refers to those data analysis systems by integrating them into a platform that allows the association of information through data transformation operations. For this analysis to be effective, it is necessary to identify and define the required quality indicators (KPIs: Key Performance Indicators) and the desired associations with them, which can be made in the BI platform. The information extraction operations can be done in the SQL queries area or when data is uploaded to the BI platform, taking into account the following factors: database queries efficiency, transformation efficiency at BI platform level, and available hardware capacity for calculations.
There is a balance between the volume of SQL queries, taking into account the complexity of the queries (number and type of joins) and the size of the files to be imported into the platform. Consideration may also be given to how new data is added to source files for the BI platform.
The applications developed for this purpose allow the analysis of KPIs by applying filters that make easier to understand the general problems of an application and identify network areas where possible degradation or shortcomings are observed.
Developing a solution for analyzing these aspects of radio measurements and comparing operators on various criteria for voice and data and application testing comes as a response to the need to be able to analyze and interpret a huge amount of data accumulated over very long periods (years) and to identify the main elements that can lead to improved network performance.
BI analysis is addressed to business managerial levels for discovery of large deficiencies without analyzing each test, satisfying the need for interaction by allocating a minimum number of parameters associated with KPIs so that analysis and decisions taken consequently have a high impact for a company benefits.
BI analysis is predominantly conducted in the telecommunications field, but the same principles can be used in other areas, and the way of working can be adapted to serve as a strong starting point in any type of analysis for which information is available at the level of a database.
In order to be able to apply the BI analysis tools, databases containing information specific to the 2G, 3G and 4G technologies (GSM, WCDMA and LTE) have been developed. Databases contain tables populated with metrics representing information similar to those collected during telecommunication measurements (data generation is done through popular spreadsheet scripts with Excel support). Multiple SQL queries for denormalizing, aggregating, and transforming data have also been developed for use as a data source in the KPIs analysis, comparison, and interpretation process.
In order to apply these data transformation queries in a format that facilitates further BI analysis, a C# application that exports the new content to a predefined area is made available to the BI environment (in text format).
The new C# content is loaded into the QLIK Sense Desktop platform, a BI development environment where another set of data transformations, (defining legends to create geo-location features for elements) is loaded in this environment. The next step is to develop specific formulas, taking into account the new available structure to obtain KPIs with expected values and views for post-denormalization analysis. Two dashboards were developed for this purpose: voice and data tests.
Configuring the development environment involves setting the connection to the location where the new content is placed, uploading these data, modifying them, creating formulas, identifying causes of value-based telecoms errors, creating a range for radio measurement values, and creating functions for geo-location of data using QLIK script syntax.
It is defined a graphical interface by creating, where necessary, maps of geo-tagged data, filters, graphs, pivot tables, pie-charts, histograms, etc. across multiple display sheets, organized to contain the information specific to a particular analysis, thereby creating an analytical context in which the information required for use in the analysis is arranged adjacent, in the form of filters or graphs, the results being easier to interpret by introducing color coding formulas.
The definition of special formulas and their application uses KPIs, and their development is carried out with a language specific to the development environment: QLIK Sense Desktop.
Part of analyzing data in a BI environment is to interpret data and various configurations. Depending on who accesses such an environment, one can configure (in the web version) different access modes and set up visible information for each user so that a manager can access general performance sheets and an RF engineer can access specific performance sheets, both of which being designed to enhance the ability of each to interpret, decide, and improve the desired performance.
Another important aspect is the low analysis time, due to the fact that the data automatically undergoes a transformation that initially had to be manually adapted. Automatic correlations that may go unnoticed, and there is also the ability to adapt and change over time. Increased efficiency comes from the fact that uncorrelated information in the database becomes correlated.
BI applications refresh after each selected criterion so that the same graph shows different information if a single operator/service is selected. The paper contains detailed BI complex analysis examples for the video player in social networks: YouTube, Facebook, Twitter, Instagram, etc.
In the BI environment, possible causes for the effects observed in the operations are performed, resulting in an easier interpretation of the results. This information can be valuable to departments such as Marketing/Sales, Network Operations Center, Optimization/Planning, etc.
Data analysis by using a BI platform is useful when data sources are not directly correlated, allowing optimized visualization and identification of KPIs performance and metrics specifically defined for fast and efficient handling of data specific to mobile telecom operators and social networks or business networks. |
