Machine Learning Predicts Soil Fertility with 92 Percent Accuracy, Helping Farmers Use Fertilizer More Efficiently
Advances in artificial intelligence are beginning to transform agriculture by helping farmers assess soil fertility more quickly and accurately. A study led by Madumere Smart Onyemaechi from Alvan Ikoku Federal University of Education, Nigeria, together with John Peter Uzoma, Ugo Chima, Bob Chile-Agada, Ihim Kingsley, and Odoemene O. Ijeoma, demonstrates that a machine learning model can predict soil fertility with an accuracy of up to 92 percent.
The research was published in 2026 in the International Journal of Sustainable Social Science (IJSSS). The researchers show that artificial intelligence can support farmers in making better decisions about fertilizer use, which can increase crop productivity while reducing environmental impact.
Challenges in Assessing Soil Fertility
Soil fertility is one of the most important factors influencing agricultural productivity. In many agrarian countries such as Nigeria, varying soil conditions often make it difficult for farmers to determine the exact nutrient requirements of crops.
Traditional soil fertility assessment methods typically rely on laboratory testing, which can be time-consuming, costly, and labor-intensive. As a result, farmers often apply fertilizer either excessively or insufficiently.
Improper fertilizer use can lead to two major problems at once: reduced crop yields and environmental damage, including soil and water pollution.
“The need for faster and more accurate methods to predict soil fertility is increasingly important in modern agricultural practices,” wrote Onyemaechi and colleagues from Alvan Ikoku Federal University of Education.
Machine Learning for Soil Fertility Prediction
To address this challenge, the research team developed a machine learning–based model designed to predict soil fertility levels. Machine learning allows computers to learn patterns from data and make predictions automatically.
The study focused on four key soil parameters that strongly influence fertility:
-Soil pH
-Nitrogen (N) levels
-Phosphorus (P) levels
-Potassium (K) levels
These nutrients play essential roles in plant growth. By analyzing the relationship between these parameters, the system can classify soil fertility into categories such as high, medium, or low.
Soil Data Collected from Five Nigerian States
The researchers collected soil samples from five major agricultural regions in Nigeria:
-Lagos
-Oyo
-Kano
-Kaduna
-Abuja
Each soil sample was analyzed based on its pH value and the concentrations of nitrogen, phosphorus, and potassium. The dataset was then used to train several machine learning models.
To ensure reliable evaluation, the dataset was divided into two parts:
-70 percent of the data was used to train the models
-30 percent of the data was used to test their predictive accuracy
This approach is commonly used in artificial intelligence research to verify that a model can perform well on new data.
Comparing Three Machine Learning Algorithms
The study evaluated three widely used machine learning algorithms in agricultural data analysis:
-Random Forest
-Support Vector Machine (SVM)
-Neural Networks
Each algorithm was trained using the Nigerian soil dataset. After training, the researchers measured the performance of each model using evaluation metrics such as accuracy, precision, and recall.
Random Forest Achieves the Highest Accuracy
The results showed clear differences in the performance of the three algorithms.
The prediction accuracy levels were:
-Random Forest: 92%
-Neural Networks: 88%
-Support Vector Machine: 85%
Among the tested models, Random Forest proved to be the most effective algorithm for predicting soil fertility based on pH, nitrogen, phosphorus, and potassium data.
According to Onyemaechi and his research team, Random Forest was better at identifying complex patterns within soil data compared to the other algorithms.
“The Random Forest model demonstrated the highest accuracy in predicting soil fertility using Nigerian soil datasets,” the researchers reported.
Supporting Data-Driven Agriculture
The findings highlight the potential of machine learning to support precision agriculture, an approach that uses data and digital technologies to optimize farming practices.
Precision agriculture enables farmers to make more informed decisions about:
-the type of fertilizer required
-the appropriate amount of fertilizer
-the optimal timing of fertilizer application
If implemented widely, machine learning–based soil fertility prediction systems could provide several key benefits.
1. Higher crop productivity
Farmers can match soil nutrients with crop needs more precisely.
2. Lower production costs
Efficient fertilizer use reduces unnecessary spending.
3. Environmental protection
Proper fertilizer application helps reduce soil and water contamination.
4. Sustainable farming systems
Digital technologies support more responsible management of agricultural resources.
In countries with large agricultural sectors such as Nigeria, these technologies could strengthen food security while improving rural economic development.
Future Development of Agricultural Technology
The researchers note that the study still has limitations, particularly regarding the size and diversity of the soil dataset used.
Future studies could expand the research by including more agricultural regions and additional soil parameters, such as organic matter content, soil moisture, or soil texture.
Further technological development may also lead to the creation of mobile applications or digital decision-support systems that can provide real-time fertilizer recommendations for farmers.
With these innovations, soil fertility prediction tools could move beyond research laboratories and become practical solutions for everyday agricultural management.
Author Profile
Madumere Smart Onyemaechi is a researcher and academic at Alvan Ikoku Federal University of Education, Nigeria, whose work focuses on the application of digital technologies and machine learning in sustainable agriculture.
The study was conducted in collaboration with several researchers from the same institution:
-John Peter Uzoma
-Ugo Chima
-Bob Chile-Agada
-Ihim Kingsley
-Odoemene O. Ijeoma
Their research interests include artificial intelligence, agricultural data analysis, and the development of technology-driven solutions to support sustainable farming systems.
Research Source
Onyemaechi, Madumere Smart; Uzoma, John Peter; Chima, Ugo; Agada, Bob Chile; Kingsley, Ihim; Ijeoma, Odoemene O. (2026).
“Machine Learning Based Soil Fertility Prediction.”
International Journal of Sustainable Social Science (IJSSS), Vol. 4, No. 1, pp. 63–66.
DOI: https://doi.org/10.59890/ijsss.v4i1.179
URL: https://aprmultitechpublisher.my.id/index.php/ijsss/index
0 Komentar