Predicting pectin performance strength using near-infrared spectroscopic data: A comparative evaluation of 1-D convolutional neural network, partial least squares, and ridge regression modeling