Aims. We aim to generate a catalogue of merging galaxies within the 5.4 sq. deg. North Ecliptic Pole over the redshift range 0.0 < z < 0.3. To do this, imaging data from the Hyper Suprime-Cam are used along with morphological parameters derived from these same data. Methods. The catalogue was generated using a hybrid approach. Two neural networks were trained to perform binary merger non-merger classifications: one for galaxies with z < 0.15 and another for 0.15 ≤ z < 0.30. Each network used the image and morphological parameters of a galaxy as input. The galaxies that were identified as merger candidates by the network were then visually checked by experts. The resulting mergers will be used to calculate the merger fraction as a function of redshift and compared with literature results. Results. We found that 86.3% of galaxy mergers at z < 0.15 and 79.0% of mergers at 0.15 ≤ z < 0.30 are expected to be correctly identified by the networks. Of the 34 264 galaxies classified by the neural networks, 10 195 were found to be merger candidates. Of these, 2109 were visually identified to be merging galaxies. We find that the merger fraction increases with redshift, consistent with literature results from observations and simulations, and that there is a mild star-formation rate enhancement in the merger population of a factor of 1.102 ± 0.084.

Aims. We present the application of a fully connected neural network (NN) for galaxy merger identification using exclusively photometric information. Our purpose is not only to test the method's efficiency, but also to understand what merger properties the NN can learn and what their physical interpretation is. Methods. We created a class-balanced training dataset of 5 860 galaxies split into mergers and non-mergers. The galaxy observations came from SDSS DR6 and were visually identified in Galaxy Zoo. The 2 930 mergers were selected from known SDSS mergers and the respective non-mergers were the closest match in both redshift and r magnitude. The NN architecture was built by testing a different number of layers with different sizes and variations of the dropout rate. We compared input spaces constructed using: the five SDSS filters: u, g, r, i, and z; combinations of bands, colours, and their errors; six magnitude types; and variations of input normalization.Results. We find that the fibre magnitude errors contribute the most to the training accuracy. Studying the parameters from which they are calculated, we show that the input space built from the sky error background in the five SDSS bands alone leads to 92.64 ± 0.15 % training accuracy. We also find that the input normalization, that is to say, how the data are presented to the NN, has a significant effect on the training performance. Conclusions. We conclude that, from all the SDSS photometric information, the sky error background is the most sensitive to merging processes. This finding is supported by an analysis of its five-band feature space by means of data visualization. Moreover, studying the plane of the g and r sky error bands shows that a decision boundary line is enough to achieve an accuracy of 91.59%.