Asset liability mismanagement has been considered a key reason for Bank failures. Banks’ loans and assets can be short-term or long-term, and their deposits and borrowings (liability) need to match the duration of loans and assets. A mismatch in the duration can lead to liquidity risk. Therefore, each bank must undertake asset-liability management as a regulatorily prudent activity. Machine learning can assist in better predicting the liability position and address volatility and seasonality in the short and long term. However, machine learning modelling techniques may not do justice to the correct prediction of the liabilities or volatility time range. This paper has explored the ensemble technique as a framework for building machine learning-based asset liability management models. It is found that ensemble techniques can deliver better predictability but address the shortcomings of standalone machine learning techniques.