Machine Learning Approaches for Forecasting the Best Microbial Strains to Alleviate Drought Impact in Agriculture
Tymoteusz Miller,
Grzegorz Mikiciuk,
Anna Kisiel,
Małgorzata Mikiciuk,
Dominika Paliwoda,
Lidia Sas-Paszt,
Danuta Cembrowska-Lech,
Adrianna Krzemińska,
Agnieszka Kozioł,
Adam Brysiewicz
Affiliations
Tymoteusz Miller
Institute of Marine and Environmental Sciences, University of Szczecin, Wąska 13, 71-415 Szczecin, Poland
Grzegorz Mikiciuk
Department of Horticulture, Faculty of Environmental Management and Agriculture, West Pomeranian University of Technology in Szczecin, Słowackiego 17, 71-434 Szczecin, Poland
Anna Kisiel
Institute of Marine and Environmental Sciences, University of Szczecin, Wąska 13, 71-415 Szczecin, Poland
Małgorzata Mikiciuk
Department of Bioengineering, Faculty of Environmental Management and Agriculture, West Pomeranian University of Technology, Słowackiego 17, 71-434 Szczecin, Poland
Dominika Paliwoda
Department of Horticulture, Faculty of Environmental Management and Agriculture, West Pomeranian University of Technology in Szczecin, Słowackiego 17, 71-434 Szczecin, Poland
Lidia Sas-Paszt
Department of Microbiology and Rhizosphere, The National Institute of Horticultural Research, Konstytucji 3 Maja 1/3, 96-100 Skierniewice, Poland
Danuta Cembrowska-Lech
Polish Society of Bioinformatics and Data Science BIODATA, Popiełuszki 4c, 71-214 Szczecin, Poland
Adrianna Krzemińska
Polish Society of Bioinformatics and Data Science BIODATA, Popiełuszki 4c, 71-214 Szczecin, Poland
Agnieszka Kozioł
Institute of Technology and Life Sciences–National Research Institute, Falenty, Hrabska Avenue 3, 05-090 Raszyn, Poland
Adam Brysiewicz
Institute of Technology and Life Sciences–National Research Institute, Falenty, Hrabska Avenue 3, 05-090 Raszyn, Poland
Drought conditions pose significant challenges to sustainable agriculture and food security. Identifying microbial strains that can mitigate drought effects is crucial to enhance crop resilience and productivity. This study presents a comprehensive comparison of several machine learning models, including Random Forest, Decision Tree, XGBoost, Support Vector Machine (SVM), and Artificial Neural Network (ANN), to predict optimal microbial strains for this purpose. Models were assessed on multiple metrics, such as accuracy, standard deviation of results, gains, total computation time, and training time per 1000 rows of data. Notably, the Gradient Boosted Trees model outperformed others in accuracy but required extensive computational resources. This underscores the balance between accuracy and computational efficiency in machine learning applications. Leveraging machine learning for selecting microbial strains signifies a leap beyond traditional methods, offering improved efficiency and efficacy. These insights hold profound implications for agriculture, especially concerning drought mitigation, thus furthering the cause of sustainable agriculture and ensuring food security.
