The Temporal Higher-order Interaction decompoSition model uses a First-Order ODE to decompose higher-order interaction sequences. It is robust and improves predictions in relation to existing methods. This addresses the limitation that traditional TF methods have in exploiting the timestamp information of data.