AN ANALYSIS OF FACTORS AFFECTING YIELD, OIL PRODUCTION RATE AND PLANT HEIGHT IN SUNFLOWERS USING SELECTED DATA MINING ALGORITHMS
S. Celik1, E. Boydak2 and Rıdvan Fırat2
1Department of Animal Science, Faculty of Agriculture, University of Bingol, Turkey.
2Department of Field Crops, Faculty of Agriculture, University of Bingol, Turkey.
Corresponding Author E-mail: senolcelik@bingol.edu.tr
ABSTRACT
In this study, yield per decare, plant length and oil production rate of sunflowers grown in the city of Bingöl, Turkey, and the elements affecting the plant length were examined using several data mining methods: Chi-square Automatic Interaction Detection (CHAID), Exhaustive CHAID, Classification and Regression Trees (CART), and artificial neural networks like Multi-Layer Perceptrons (MLP) and Radial Basis Function networks (RBF). Yield per decare was affected by oil production, the 1000 grain ratio, and the kernel percentage. Oil production rate was affected by the kernel percentage, plant length, 1000 grain ratio and yield per decare. Plant length was affected by the kernel percentage. These effects were significant. According to the Exhaustive CHAID algorithm used to predict yield per decare, the Pearson correlation coefficient (r), coefficient of determination (R2), adjusted coefficient of determination (Adj. R2), RRMSE, RAE and SD ratio were recorded as 0.915, 0.837, 0.825, 0.00025, 0.060 and 0.403, respectively. The largest average yield per decare (297.643 kg) was obtained from the subgroup consisting of plants whose oil production rate was ≤ 28.040 and grain ratio was > 83.800. According to the CART algorithm used to predict oil production rate (%), r, R2, Adj. R2, RRMSE, RAE and SD ratio, the results were calculated to be 0.988, 0.976, 0.975, 0.00045, 0.0021 and 0.155, respectively. The largest average oil production rate was obtained from the sub-group consisting of plants whose kernel percentage was > 62.500, whose plant length was >127.500 cm and whose kernel percentage was > 72.250. According to the Exhaustive CHAID algorithm used to predict plant length, r, R2, Adj., R2, RRMSE, RAE and SD ratio, these were calculated to be 0.942, 0.887, 0.883, 0.00021, 0.029 and 0.335, respectively. The longest average plant length (166.600 cm) was obtained from the subgroup consisting of plants where 69.500 ≤ kernel percentage < 70.800.
As a result, the use of Exhaustive CHAID and CART algorithms can be employed in field crops studies to predict some plant characteristics of sunflowers.
Keywords: Data mining, sunflower, plant characteristics.
|