Food quality analysis methods play an important role in food control, safety and logistics. Bioimpedance spectra were obtained for frozen and thawed chicken meat samples using the AD5933 development kit from Texas Instruments. These measurements were fitted to the Cole Cole Model using the least squares method. A convolutional neural network - CNN, was designed and trained to classify the freezing process of chicken breast samples using Cole Cole model parameters for the experimental dispersion. A training data expansion method for the CNN is proposed. The advantages of using CNN over geometric methods for food classification using electrical bioimpedance spectroscopy data are discussed. The results show a maximum deviation of 5% over the CNN output in predicting the history of a test data set.