As a unique probe, the precision measurement of pp solar neutrinos is important for studying the sun’s energy mechanism as it enables monitoring the thermodynamic equilibrium and studying neutrino oscillations in the vacuum-dominated region. For a large-scale liquid scintillator detector, a bottleneck for pp solar neutrino detection is the pile-up events of intrinsic 14141414145×10-181414C decay. This paper presents a few approaches to discriminating between pp solar neutrinos and 14141414145×10-181414C pile-up events by considering the differences in their time and spatial distributions. In this study, a Geant4-based Monte Carlo simulation is conducted. Multivariate analysis and deep learning technology are adopted to investigate the capability of 14141414145×10-181414C pile-up reduction. The BDTG (boosted decision trees with gradient boosting) model and VGG network demonstrate good performance in discriminating pp solar neutrinos and 14141414145×10-181414C double pile-up events. Under the 14141414145×10-181414C concentration assumption of 14141414145×10-181414 g/g, the signal significance can achieve 10.3 and 15.6 using the statistics of only one day. In this case, the signal efficiency for discrimination using the BDTG model while rejecting 99.18% 14141414145×10-181414C double pile-up events is 51.1%, and that for the case where the VGG network is used while rejecting 99.81% of the 14141414145×10-181414C double pile-up events is 42.7%.