Data visualization dashboards provide an e cient approach that helps to improve the ability to understand the information behind complex databases. It is possible with such tools to create new insights, to represent keys indicators of the activity, to communicate (in real-time) snapshots of the state of the work. In this paper, we present a visual analytics platform created for the exploration and analysis of performance data on entrance tests taken by Italian students when entering the university career. The data is provided by CISIA (Consorzio Interuniversitario Sistemi Integrati per l'Accesso), a non-pro t consortium formed exclusively by public universities. With this platform, it is possible to explore the performance of the students along di erent dimensions, such as gender, high school of provenience, type of test and so on.
The e ective graphical presentation of information is an essential skill in most
scienti c disciplines. Visualization is intended to clearly convey and communicate
information through graphical means, enabling everyone to comprehend data in a
much more explicit way[
Copyright c 2019 for the individual papers by the papers' authors. Copying permitted for private and academic purposes. This volume is published and copyrighted by its editors. SEBD 2019, June 16-19, 2019, Castiglione della Pescaia, Italy.
A proper way to describe graphically a complex phenomenon is through
the use of dashboards. As stated in [
In this paper we present a visual analytic platform created for the exploration and analysis of performance data on entry tests taken by Italian students before entering the university career. The data is provided by CISIA4 (Consorzio Interuniversitario Sistemi Integrati per l'Accesso), a non-pro t consortium formed by public universities. Currently, CISIA consortium counts 45 Universities and the Conferences of Engineering, Architecture and Sciences among the consortium members: the CUIA - Italian University Architecture Conference, the CopI - Conference for Engineering and Con.Scienze - National Conference of Presidents and Structure Directors University of Science and Technology.
The Consortium is open to the participation of all Italian universities; among the di erent statutory purposes, the main is to organize and coordinate the orientation activities for the access to the universities. Entrance test organized by CISIA serve two di erent purposes: { for students enrolling the test, it provides a self-assessment of their preparation and aptitude to undertake the chosen eld of studies; { for the faculties and departments, the tests gives a view of the actual knowledge of the students, and let the management to compare it with the minimum skills and capabilities requirements in order to prepare speci c orientation and training activities.
For those course with a restricted number of students, the tests are used as screening and ranking tools. CISIA tests are currently available for six areas: Engineering, Economics, Pharmacy, Sciences, Humanities, and Agriculture. 2 2.1
CISIA Online Test (acronym TOLC) is a tool for orientation and assessment of the knowledge required for access to the Study Programs of Italian Universities, which can be used to select students for access, provided on a computerized platform. TOLC is an individual test, di erent from student to student, automatically composed for each student by a software. The software follows a set of recipes to guarantee that all the tests generated are equivalent in terms of the level of di culty. The objective of the platform proposed in this paper is 4 http://www.cisiaonline.it to provide an analytic framework where all the performances on each test may be analyzed. Historical data are maintained since 2012, and they register many dimensions from each test: the number of correct answers, high school education of the student attending the test, the geographical origin, the category of each question in the test. Thus the platform should allow an e cient exploration of the many dimensions and should produce an accessible interface where the observer may interact with the data. During the design phase, we identi ed two kind of users: public visitor of the website of the consortium, and the site managers of the department that organize the tests. The two pro les are managed through a similar web interface, whereas the users with an authorized account may access more detailed data, for example analyzing the performance of the tests organized in her own site. 2.2
Accordingly to Sarikaya et al. [
A schematic representation of the architecture of the CISIAViz platform is presented in Fig.1. The database layer is organized with two levels. The production database used to manage the tests is kept separated from the CISIAViz platform. An ETL (Extract, Transform, and Load) procedure is scheduled to update a staging database for the visualization. During this phase, sensitive information is removed.
The database is running over an instance of MySQL, the same DBMS adopted for the rest of the TOLC system. The data warehouse schema in Fig.2 shows the relations implemented within the DB. The entity Test is the main table containing all the results of each test. Since the sections for all the possible tests are limited, we decided to store them into the eld within this table. The test Fig. 1. Web-App schema: Starting from CISIA database, through a process of Extract, Transform, Load (ETL), we build the data warehouse (DB viz). The data warehouse contains only the data that is of interest to the CISIAViz platform. During the ETL phase, new dimensions are calculated (such as the ranking of students by type of test) and some queries are optimized by materializing data (including redundant ones) retrieved from di erent tables to speed up the subsequent reading phase via web API. provided in a di erent area of interests may have di erent compositions. When a test does not contain a speci c section, the corresponding value in the table will be NULL. On one side, this implementation makes the schema of the DB not exible to accommodate new sections. On the other hand, this allows much faster access during the visualization, as it does not need to create too many joins between the tables.
The relevant dimensions of the star schema are Geo, the geographic dimension, and Time, the time dimension: the rst one is used to explore the origin of the students applying for the test, and the second one allows us to select the time interval to aggregate the tests. Although it is possible to aggregate for an arbitrary temporal interval, to guarantee a su cient aggregation for sensitive information, the time selection is restricted at the year level.
The access to the database is managed by an application server running a web API implementing the endpoints to access the data. All the interaction of the user with the visualization dashboard will be translated into the corresponding call from the web application. The access to the web API is managed through HTTP GET calls. The authentication, implemented with an SSO system (Single Sign-On) using an IDP (Identi cation Provider) of the consortium, is performed with HTTP-POST methods. When the user changes the preferences on the dimensions, the web application generates a set of call to the Web API to retrieve the updated data. The application server is implemented in Python, leveraging the Flask library to handle the implementation of the endpoints. The web application running in the front end is a single page application running in Javascript. The dashboard is implemented using the state of the art libraries for managing of user interfaces: D3.js to create custom visualization; Vue.js to manage the events and selection of the user and to make the internal data representation reactive to new updates; Plot.ly to construct high-level charts using con guration descriptors automatically generated by data. 2.3
The majority of the endpoints of the Web API share a common parameter structure. This design choice allows us to make the front end exible by composing the set of parameters automatically to be provided to the server. All the selection a user can make on the web interface are projected into a sequence of parameter values that are sent to the web server. All calls that depend on this lter have the following scheme: @ app.route('/api/function_rest/<year>/<area_interest>/<region>/<quantile>') where <year> and <area_interest> are the only mandatory parameters. The parameters re ect the possible combination of the dimensions the user may explore: <years> Selection of the year in a range between 2012 and the current year <interest> Select the type of test. Currently, TOLC-I (engineering), TOLC-E (economics), TOLC-F (pharmacy) can be selected. <region> Selection of the region of origin of the candidates to the tests. In case of missed selection (default status) the value all is used. <quantile> during the ETL phase, each candidate is associated with a percentile ranking positioning with respect to the score of all the candidates who carried out the test. This lter allows selecting candidates with values higher than a certain percentile. To simplify the exploration for non-expert users, the CISIAViz platform allows selecting only the \best scores", i.e. the score of the top 10th percentile of tests in the DB for the current combination of year and area of interest. 2.4
The CISIAViz platform is implemented as a single-page web application. The platform is deployed within the website of the consortium5. 5 http://www.cisiaonline.it/area-tematica-cisia/report/
The main script is the entry point to the application and is responsible for aligning the selection on the interface with the data to the DB, by querying the Web API with new data when needed. Here the event handling of the application is orchestrated and it updates all graphs when new data arrives. Every chart is handled by a general class module chart, whose input is transformed accordingly to the type of gure to show using a dedicated data transformer. The con guration and the mapping of transformers to charts are described into the chart con g. module. The modules lter, toolbar, event handler take care of sending the requests to the server.
The web application is composed of a central column showing all the statistics, and a sidebar where the user can manage to insert the constraint for the dimensions to explore: year, area of interest, region of origin, best scores.
The central column is organized into sections, each of them focusing on a speci c aspect of the data: { Volumes : the rst section shows the number of tests executed with reference to the actual selection, presenting the partial sums for sub-periods of the year. { Scores: this section shows the average score per section in the selected tests and the distribution of each section. { Repetition of tests : here we show how the score varies for those students that repeated the test multiple times. { Region of origin: here we cross lter both the geographical origin and the high school diploma of the students attending the test. { Historycal data: here we show the trends of the test along the years. This section is not a ected by the year selection in the sidebar since shows always the statistics of the last ve years. 3
To demonstrate the e cacy of the web application to deliver high-level knowledge extracted from the data, we show the capabilities of the dashboard through a use case of a person interested in the performance of students. For example, let us consider a journalist interested in documenting the skills of students when entering university: a study of how many individuals take the entry test, how they behave in terms of performance and the presence of di erences between types of students, could be an engaging topic for an article on young people. Leveraging the dashboard for the di erent areas of interest, it is possible to get information on the distribution, among gender, of students taking the tests, as shown in gure 3.
From the gure, it can be seen how, also in the testing phase, the women are underrepresented especially among STEM degree (in this case Engineering). The higher representation of female students is found for Economics and Pharmacy.
It is also possible to compare the performance between male and female students in a di erent section of the test: for this graph (but also for others) there are local selectors with which selecting the female, the male or the total population who took a speci c test (TOLC). In gure 4 it is shown the distribution of scores among male (top image) and female (bottom image) students. No signi cant di erences can be found between the two genders, except for the results obtained in mathematics, where male students perform better than females. In the gure 4 is reported the distribution of scores for the test for Economics in 2018, but obviously, the same information can be obtained for the other areas (TOLC) and also the trend among years can be seen changing the selector for the years. Another possible comparison can be made among di erent Italian regions. In gure 5 the average scores obtained to the tests is reported for each region. The results reported in the gure are related to the TOLC in Economics, but the same results can be obtained for Pharmacy or Engineering. In this case average score for each section of the TOLC-Economics of 2018 divided by region and school of origin (all in this speci c case). Only regions with a signi cant number of tests performed are shown. Furthermore, a comparison among di erent high schools can be made. The detailed average score has been calculated for scienti c high school, classical high school, technical institute, technical institute for surveyors, commercial technical institute, professional institute and magistral and pedagogical institute.
The dashboard, therefore, allows you to get a picture of the situation of the students who decide to follow the course of study: a journalist interested in this topic and the di erences between students with diverging characteristics (both gender or region of origin), can create his article guided by the data and statistics shown on the portal.
In this paper we present a visual analytic platform to explore the performances of students when applying for entry tests to the universities. The platform was developed using the state-of-the-art of the technologies for web developing. The visual design of the dashboard was developed to support the user through a set of steps to select sub-dimension of the data. The toolbox which guides the composition of the visual platform is organized in steps and the user may select the relevant parameters to start the exploration. More speci c selections, like gender selection or the selection of the high school diploma of origin, are located close to the speci c corresponding chart. The system has been deployed and it is used e ciently both by non-expert users and domain experts (with login-based access).