Detection of oil yield from oil shale based on near-infrared spectroscopy combined with wavelet transform and least squares support vector machines

In order to improve the prediction accuracy for the low reflectance spectra of the oil shale samples, the detection of the oil yield from the oil shale was established by using the near-infrared spectroscopy combined with the wavelet transform method and the least squares support vector machines algorithm. The calibration and prediction sets of the oil shale samples were portioned by the Kennard-Stone algorithm. The wavelet transform method was used for denoising and compressing signals. Ant colony optimization algorithm was employed for variable selection. The results show that without using ant colony optimization algorithm, the model based on continuous wavelet transform method is optimal, and the wavelet function and scale is db9 and 7 respectively. Then determination coefficient and root mean square error prediction is 0.94 and 0.38. Using ant colony optimization algorithm, the model based on the discrete wavelet transform method is the best, and determination coefficient and root mean square error prediction of prediction set is 0.96 and 0.38 respectively. Therefore, the proposed solution is suitable for determining the oil yield of oil shale and helpful for other rock samples of the low reflectance spectra.