Real World Dataset for Maize Lethal Necrosis Disease Classification in Kenya Incorporating Visual, Texture, and Spatial Features under Class Imbalance

1. Data DescriptionThis dataset contains structured observations for the detection and classification of Maize Lethal Necrosis Disease (MLND) collected from major maize growing regions in Kenya. The dataset comprises 7,892 samples, where each observation represents a maize leaf characterized using interpretable visual, texture, and spatial features derived from field assessments and image-based analysis.

The primary objective of the dataset is to support the development and evaluation of machine learning models for MLND detection under realistic agricultural conditions, including class imbalance. The target variable, Disease_Label, is binary, indicating whether a maize leaf is healthy (0) or infected with MLND (1).

The dataset includes one categorical variable (County), representing the geographic origin of the sample, and eleven continuous features scaled between 0 and 1. These features capture key disease-related characteristics, including chlorosis intensity, green to yellow tissue ratio, necrotic area, texture roughness, patchiness, lesion distribution, elongated streak patterns, irregular lesions, infected distribution, boundary sharpness, and transition smoothness. Together, these variables provide a comprehensive representation of disease symptoms observable on maize leaves.

This dataset is particularly valuable for research on interpretable machine learning, class imbalance handling, and agricultural decision support systems. Unlike many existing datasets that focus on raw images, this dataset provides engineered features that enable transparent and computationally efficient modeling, making it suitable for deployment in resource constrained environments.

The dataset can be used for classification, model benchmarking, explainable artificial intelligence studies, and comparative algorithm evaluation. It is intended to support reproducible research and contribute to the advancement of data driven approaches for plant disease detection and food security.

2. Target Variable (Dependent Variable)

2.1 Disease_Label

·         Description: Indicates the health status of maize leaves

·         Type: Categorical (Binary)

·         Values:

o   0 = Healthy maize leaf

o   1 = Maize Lethal Necrosis Disease (MLND) infected leaf

·         Role: Dependent variable (prediction target)

3. Independent Variables (Features)

A. Location Feature

3.1 County

·         Description: Geographic location where maize sample is assumed to originate

·         Type: Categorical

·         Values:

o   Trans Nzoia

o   Uasin Gishu

o   Nakuru

o   Narok

o   Kakamega

o   Bungoma

B. Visual Features (Color-Based)

3.2 Chlorosis_Intensity

·         Description: Level of yellowing in maize leaf due to disease

·         Type: Continuous (0–1)

·         Interpretation: Higher values indicate stronger yellowing symptoms

3.3 Green_Yellow_Ratio

·         Description: Ratio of healthy green tissue to yellow (chlorotic) tissue

·         Type: Continuous (0–1)

·         Interpretation: Lower values suggest higher disease severity

3.4 Necrotic_Area

·         Description: Proportion of brown or dead tissue on leaf

·         Type: Continuous (0–1)

·         Interpretation: Higher values indicate advanced infection

C. Texture Features

3.5 Texture_Roughness

·         Description: Measures irregularity of leaf surface texture

·         Type: Continuous (0–1)

·         Interpretation: Higher values indicate more damaged/irregular surface

3.6 Patchiness

·         Description: Degree of irregular infected patches on leaf surface

·         Type: Continuous (0–1)

3.7 Lesion_Distribution

·         Description: Spread pattern of lesions across the leaf surface

·         Type: Continuous (0–1)

D. Shape & Pattern Features

3.8 Elongated_Streaks

·         Description: Presence of elongated yellow streak patterns typical of MLND

·         Type: Continuous (0–1)

3.9 Irregular_Lesions

·         Description: Irregular shape of disease spots on leaf surface

·         Type: Continuous (0–1)

3.10 Infected_Distribution

·         Description: Overall distribution intensity of infected regions across leaf

·         Type: Continuous (0–1)

E. Edge Features

3.11 Boundary_Sharpness

·         Description: Sharpness of boundaries between healthy and infected regions

·         Type: Continuous (0–1)

·         Interpretation: Higher values indicate clearer disease demarcation

3.12 Transition_Smoothness

·         Description: Smoothness of transition between healthy and diseased tissue

·         Type: Continuous (0–1)

·         Interpretation: Higher values indicate gradual change in tissue condition