453 Machine learning techniques for inspection data analysis Rajko Bulatovic * , Zora Konjovic ** , Aleksandar Ivic ***, Djordje Obradovic *** * Provincial Secretariat for Health, Vojvodina, Serbia ** Singidunum University, Belgrade, Serbia *** Faculty of Technical Sciences, University of Novi Sad, Serbia rajko.bulatovic@vojvodina.gov.rs, zkonjovic@singidunum.ac.rs, ivicaca@gmail.com, obrad@uns.ac.rs, Abstract - The unique data set was created by collecting data about sanitary inspection control on the territory of AP Vojvodina during the period of two years. The total number of multi-dimensional data records is 28,403. The Open Data with complete details is published on web site of Provincial Health Secretariat in the section Documents – Data, available for a bulk download. Data comprise temporal, spatial and categorical components and as such are highly suitable for a variety of analyses by means of machine learning techniques, especially neural networks. In this paper examples of linear regression and neural networks applications to analysis of the data are presented. The obtained results can be used for improving daily tasks like estimating inspection control workload, and alike. With the adoption of the Law on Inspection Control (in further text referred as Law) in 2015, Serbian Parliament opened the way for a paradigm shift in government inspection services. In fact, only with the adoption of this Law the preconditions for fundamental reform of governmental inspection services were established. The adoption of the Law was preceded by analysis, study visits to countries that cultivate a good inspection practice, and consulting local and foreign experts’ consultation in the process of passing the Law. The result was the Law that was positively evaluated not only by non-governmental organizations but also by the European Commission. One of the mechanisms which create the preconditions for the lawful and proper operation of the inspection managers and inspectors are proper information systems and software solutions that enable the transparent work of inspections, harmonization of inspection practices, timely access to data, case monitoring, risk evaluation, and corruption prevention. [1] An important new aspect of the Law is regulation of the transparency of governmental inspection services. The Law defines, but does not restrict, activities aimed at establishing transparency, e.g. publication of inspection control plans and reports , inspection checklists, best and worse rated subjects of inspections (white and black lists), and current regulation concerning inspections on public internet sites. The practice of publishing required information is not widespread at the moment among inspections in Serbia, mostly due to the lack of technical capacity. This paper proposes publishing additional non- mandatory data extracted from the database and/or derived from the raw data. Furthermore, the conducted data analysis could be the base for resource planning in future inspection supervisions. In 2011 the Provincial Government of Vojvodina enjoined the Provincial Department of Health to prepare the tender documents for the implementation of the software for sanitary control. The goal was a software system aimed at improvement of the entire inspection process. More precisely, in order to achieve this global goal, the software system should provide for:  Automated creation of documents with electronic archive functionality.  Efficient creation and maintenance of the following up to date electronic databases: o Registry of objects under sanitary control, o Registry of persons under sanitary control, o Registry of relevant regulations and legislation used in the control.  Monitoring of sanitary and hygiene conditions in facilities under sanitary supervision.  Significant improvement of the functions of planning inspection control, autonomous processing of reports, monitoring of inspection process (current outcome and legal assistance to the inspector from start to the end of the control process).  Assisting in establishing a uniform inspection methodology and procedures.  Equal treatment for all subjects under supervision.  Creating the conditions for a comfortable, efficient, precise, legally safe operation of inspectors.  Supervision of the work and expertise of inspectors.  Internal licensing (periodic proficiency testing) for inspectors.  Preventing illicit relations, corruption, and other adverse consequences.  Raising transparency of inspection by publishing information regarding regulations, practices, lessons about appeals, answers to frequently asked questions, educational materials and the like. Good international practice in the organization of the inspection that is sublimated in the recommendations of the World Bank presented the study: “Good Practices for Business Inspections - Guidelines for Reformers, World Bank Group / Small and Medium Enterprise Department – 2006” which was the starting basis for requirements specification. In 2012 the team from the Faculty of Technical Sciences in Novi Sad implemented the software for sanitary control with requested features. The software is in the production phase since January 2013 on the territory of AP Vojvodina. 7th International Conference on Information Society and Technology ICIST 2017