We present a novel approach to estimate gas hold-up in bubble columns using wall pressure fluctuations – without requiring knowledge of operating flow regime, column configuration or physical properties. The approach uses machine learning and pressure fluctuations acquired from a wall mounted pressure sensor. The approach is validated by carrying out experiments with different physical properties of liquid (deionised water, deionised water – alcohol [ethanol, propanol and butanol] mixture up to 2 % alcohol, deionised water – glycerine (30 %), tap water). The majority of the data was obtained with 0.1 m diameter bubble column. Data with tap water in this column was obtained for four different spargers. Data with tap water was also obtained for 0.15 m and 0.33 m diameter bubble columns. Gas velocity was varied up to 0.1 m/s. The covered physical properties, sparger configurations, column diameter and gas velocities span different flow regimes. An artificial neural network (ANN) based model was developed to relate characteristics of wall pressure fluctuations and superficial gas velocity to the gas hold-up. The approach was validated by comparing predicted results with the experimental data unseen by the ANN model. The approach demonstrates the feasibility of using wall pressure fluctuations with machine learning to estimate key characteristics without prior knowledge of column configuration, flow regimes or physical properties.