Data-Driven Intelligent Crop Advisory System Using Machine Learning Approaches for Kenya’s Central Highlands

다국어 초록 (Multilingual Abstract)

Farming is key in Kenya, but small scale farmers in the Central Highlands deal with poor soils and uncertain weather. Extension services are stretched thin, and digital tools often give generic advice that doesn’t match local needs. Because of this, farmers get lower harvests and stay vulnerable. This study designs a crop advisory system that uses machine learning to give farmers advice that is simple,
flexible, and local. It follows the Information System Success Model, making sure farmers are part of the process and that their feedback helps build trust. This research links to Kenya’s Agricultural Sector Transformation and Growth Strategy (ASTGS), Bottom-Up Economic Transformation Agenda (BETA), and the global Sustainable Development Goals SDG’s, aiming to prove the value of crop planning that is local and adaptable. The dataset had 2,196 entries, with 7 input features and 21 different crop categories. Machine learning was used to make predictions and adjust them as conditions changed. Farmers’ feedback made the system even better, improving both accuracy and usefulness. In testing, the system was checked for accuracy, usability, and policy alignment. The findings revealed that machine learning worked better than traditional advisory approaches, especially in predicting yields, helping farmers adapt to climate change, and buildingiitrust. This proves that farmer-centered machine learning can play a big role in shaping agriculture in Sub-Saharan Africa and supporting Kenya’s policies.

목차 (Table of Contents)

• 1. Introduction 1
• 1. Problem Statement 7
• 2. Objectives and goals 10
• 3. Significance of the Study 12
• 2. Literature Review 16
• 1. Kenya: Agriculture in brief 18
• 1.1 Structure and Scope of Agriculture in Kenya 20
• 1.2 Land Use Trends and Production 21
• 1.3 Food Policy, Nutrition and Digital Opportunities 22
• 1.4 Environmental and climatic vulnerabilities 25
• 1.5 ICT in Agriculture in Kenya 29
• 2. Plant Nutrients and Soil Requirements in Crop Recommendation Systems 43
• 3. Data Preprocessing and model design 46
• 4. Machine Learning Techniques 48
• 5. Deep Learning Techniques 53
• 6. Model Evaluation and Criteria 58
• 7. Crop Recommendation Systems 61
• 8. Theoretical Background of the study 67
• 9. Research Gap 92
• 10. Research Framework 93
• 11. Research Hypotheses 95
• 3. Materials and Methods 108
• 1. Research design 108
• 2. Methodology 110
• 2.1 Area of Study 110
• 2.2 Data Collection 111
• 2.3 Data Preprocessing 112
• 2.4 Explorative Data Analysis 113
• 2.5 System Design 136
• 2.6 Deployment Architecture 148
• 4. Results 151
• 1. Performance Evaluation 151
• 2. Confusion Matrix 153
• 3. Radar Chart Analysis: Multi-Metric Model Comparison 166
• 5. Discussion 171
• 1. Major Findings in Relation to Literature 173
• 1.1 Theoretical and Practical Implications 174
• 1.2 Conformity with Kenya's Strategic Policy and Development Objectives 176
• 1.3 Socio-Technical Consequences 179
• 1.4 Ethical concerns 179
• 2. Limitations 184
• 3. Future Work 186
• 6. Conclusion and Recommendations 189
• 7. References 197
• Appendix 1- Data Request Letters 234
• Appendix 2 – ICT Policy 236
• 1. Executive Summary 236
• 2. Introduction 237
• 3. Literature Review 239
• 4. Kenya’s Current Situation 245
• 5. Policy Analysis 246
• 6. Recommended Policy 246
• 7. Negotiation Regulation, Legislation, Enforcement and Evaluation 258
• 8. Final Policy and Discussion 259
• 9. Conclusion 261
• 10. References 263