The caldera of Campi Flegrei (Southern Italy) has shown an intensification in terms of seismic activity, ground deformation and geochemical parameters during the current unrest. In this work we considered the increase in seismicity, in relation to the number and amplitude of earthquakes. We chose 63 earthquakes recorded by the CSOB seismic station in the period between January 3, 2023 and April 30, 2024, which also includes the occurrence of seismic swarms. Then, we performed an unsupervised analysis of these earthquakes to identify possible clusters of events by using the Self-Organizing Map (SOM) neural network, widely used in previous applications and able to group data by exploiting intrinsic measures of similarity in the data itself. Before applying the SOM, the data were preprocessed in order to encode them through features that would characterize them in a unique way. Three types of experiments were carried out: in the first, the SOM was applied directly on the seismograms of the signals; in the second, the Linear Predictive Coding (LPC) technique was used to extract the envelope of the signals in the frequency domain; in the third, a waveform parametrization was added to the LPC coefficients in order to also have characteristics of the signals in the time domain. The best results were obtained with the third experiment, in which the SOM clustering identified two main families of events that could be related to different physical processes describing the seismic activity of the caldera.