Invasive species can have irreversible effects on the ecosystems into which they are introduced, which is why the most effective way to deal with them is to prevent invasion. Essential for effective prevention are understanding the invasion process and quantifying the risk. Species distribution modeling (SDM), which predicts suitability for species based on occurrence and environmental data, is frequently employed for these objectives. However, such use of SDM has sometimes been questioned because SDM often overlooks anthropogenic introduction pressure. This study models the invasion process in two distinct aspects. The first is local propagule pressure (LPP), which gauges the likelihood or intensity of introduction of non-native species through anthropogenic activities. The second is habitat suitability index (HSI), which indicates whether the local environmental conditions are conducive to the establishment of non-native species. We focused on two non-native strictly freshwater shrimps, Neocaridina davidi and Palaemon sinensis, which have been introduced throughout Japan. Their dispersal is restricted within a water system and anthropogenic introduction pressure would play a significant role in their invasion. Employing machine learning on extensive spatio-temporal occurrence/non-occurrence data for the two non-native shrimps, we developed LPP and HIS models. Integration of these models enabled the prediction of newly invaded locations in recent years that were not included in the model training datasets. Furthermore, employing the explainable AI (XAI) approach allowed us to gain insights into the differing spreading and establishment processes of these two non-native shrimps. This highlights the efficacy of both LPP and HIS in capturing different aspects of the invasion process and predicting locations at high risk of invasion. Our modeling approach can be applied to various invasive species for which spatio-temporal occurrence/non-occurrence data (such as monitoring data) exist.