Abstract： To improve the model applicability of calibration model built from spectroscopy, the model maintenance is needed when the prediction precision of original model is poor for predicting new samples.The discriminant analysis (DA) was used to build transgenic tomato parent sample discriminant calibration model to predict sample ripeness with same batch at same ripeness stage, different batches at same ripeness stage and different batches at different ripeness stages. To improve model applicability, model update by adding new typical samples into original model and absorbance modification were adopted. The results indicate that calibration model can be employed to predict samples with same batch at same ripeness stage. The prediction performance of transgenic samples with different batches at same ripeness stage can be improved by model update, and the prediction performance of parent samples can be improved by absorbance modification. For predicting transgenic tomatoes and parent samples with same batch at different ripeness stage, the efficiency of absorbance modification is better than that of model update.
2012, Vol. 33 
