Primary productivity is in the foundation of farming communities. Therefore, much effort is invested in understanding the factors that influence the primary productivity potential of different soils. The International Long-Term Ecological Research (ILTER) is a network that enables valuable comparisons of data in understanding environmental change. In this study, we investigate three ILTER cropland sites and one long-term field experiment (LTE) outside of the ILTER network. The focus is on the influence of different management practices (tillage, crop residue incorporation, and compost amendments) on primary productivity. Data mining analyses of the experimental data were carried out in order to investigate trends in the productivity data. We generated predictive models that identify the influential factors that govern primary productivity. The data mining models achieved very high predictive performance for each of the sites. The preceding crop and the crop of the current year were crucial for primary productivity in the tillage LTE and compost LTE, respectively. For both crop residue incorporation LTEs, plant-available Mg affected productivity the most, followed by properties such as soil pH, soil organic matter, and the crop residue management. The results obtained by data mining are in line with previous studies and enhance our knowledge about the driving forces of primary productivity in arable systems. Hence, the models are considered very suitable and reliable for predicting the primary productivity at ILTER sites in the future. They may also encourage researcher-farmer-advisor-stakeholder interaction, and thus create enabling environment for cooperation for further research around these ILTER sites.
F.02 Acquisition of new scientific knowledge
COBISS.SI-ID: 31437607We have developed a decision support system (DSS) to aid farm advisors in providing farmers with advice on how to reduce the risk of water pollution when using pesticides in agriculture. To develop the DSS, we used Multi Criteria Decision Analysis, as implemented into the DEX (Decision EXpert) integrative methodology of building qualitative multi-attribute decision models with the DEXi modelling tool. We captured expert knowledge about the application of different crop and soil management practices for the reduction and elimination of water pollution with pesticides, as well as about meteorological conditions and soil properties. We also integrated knowledge derived by mining data collected at the experimental farming site of La Jailliere in France. The developed DSS comprises risk assessment and risk management modules and was validated at the experimental center La Jailliere (a reference center for the European Commission FOCUS working group). To facilitate the use of the DSS by end-users, we built a web application.
F.15 Development of a new information system/databases
COBISS.SI-ID: 31574055We used the random forests algorithm to build models for assessing the risk of water pollution by pesticides in field-drained outflow water. The models are based on empirical data from field trials at the La Jailliere experimental site in France. To address the problems of the imbalanced class distribution in the data, cost-sensitive learning and different measures of predictive performance were used. Despite the highly imbalanced distribution between risky and not-risky application events, we managed to build predictive models that make reliable predictions.
F.01 Acquisition of new practical knowledge, information and skills
COBISS.SI-ID: 31356967We organized a Summer school on mining big and complex data (http://maestra-project.eu/school/). The school included lectures on predictive modelling methods for big and complex data. More specifically, the lectures presented methods handling the following complexity aspects: (a) structured data as input or output of the prediction process, (b) very large/massive datasets, with many examples and/or many input/output dimensions, where data may be streaming at high rates, (c) incompletely/partially labelled data, and (d) data placed in a spatio-temporal or network context. The applicability and the potential of the presented methods was demonstrated on showcases from environmental sciences (including agriculture), molecular biology, sensor networks, multimedia, and social networks. More specifically, on the topic of environmental applications, several studies performed within the project were presented, such as modelling soil functions and predicting the quality of agricultural waters. The lectures were given by world leading researchers and experts in machine learning and data mining, as well as experts from the application domains. The school was attended by approximately 100 participants. The lectures were recorded and are available at videolectures.net.
B.01 Organiser of a scientific meeting
COBISS.SI-ID: 30338855Professor Džeroski was the general chair of the conference ECML PKDD 2017, the most prestigious European event in the areas of machine learning, data mining, and more generally data science. Even though European, this conference has a highly international character and a long tradition, dating back to 1986: ECML PKDD 2017 was the 28th European Conference on Machine Learning and the 21st European Conference on Principles and Practice of Knowledge Discovery in Databases. Many members of the project team (e.g., dr. Dragi Kocev, dr. Panče Panov) played key roles in the organization of the conference. The conference took place in Skopje, Macedonia and attracted more than 600 participants. Its proceedings are published in three volumes of the series Lecture Notes in Artificial Intelligence by Springer.
B.01 Organiser of a scientific meeting
COBISS.SI-ID: 31153703