Crying is the only mode of communication for babies to share information with the surrounding environment. Expert knowledge is required to analyze the audio signals in pre-processing and extract the features from it. Deep learning doesn't require much pre-processing and it automatically extracts the crucial features directly from the data. This paper presents the deep learning based feature extraction and machine learning based classification approach in infant cry classification and compares the various machine learning algorithms in infant cry classification. The audio signal of length 4 seconds data was converted into spectrogram image which was taken as input, deep convolutional neural network extract the features based on that the data were classified using various machine learning algorithms such as SVM, Naïve Bayes and KNN then compare their performance with Bayesian Hyper-parameter optimization technique. The experimental result shows that SVM provides better performance than K Nearest Neighbor and Naïve Bayes in the classification of infants hunger, pain and sleepy cries. © 2020 IEEE.