In this work, feasibility of flow pattern and oil holdup prediction for vertical upward oil–water two–phase flow using pressure fluctuation signals was experimentally investigated. Water and diesel fuel were selected as immiscible liquids. Oil holdup was measured by Quick Closing Valve (QCV) technique, and five flow patterns were identified using high speed photography through a transparent test section with Inner Diameter (ID) of 0.0254 m. The observed flow patterns were Dispersed Oil in Water (D O/W), Dispersed Water in Oil (D W/O), Transition Flow (TF), Very Fine Dispersed Oil in Water (VFD O/W) and a new flow pattern called Dispersed Oil Slug & Water in Water (D OS& W/W). The pressure fluctuation signals were also measured by a static pressure sensor and decomposed at five levels using wavelet transform. Then, standard deviation values of decomposition levels were used as input parameters of a Probabilistic Neural Network (PNN) to train the network for predicting the flow patterns. In addition, some considered numerical values for actual flow patterns together with signal energy value of each level were used as input parameters of a MultiLayer Perceptron (MLP) network to estimate the oil holdup. The results indicated good accuracy for recognition of the flow patterns (accuracy of 100% and 95.8% for training data and testing data, respectively) and oil holdup (AAPE=9.6%, R=0.984 for training data and AAPE=8.07%, R=0.99 for testing data).