A glass fibre reinforced thermoplastic textile composite material is characterized in terms of its strain rate dependent 3D shear deformation behaviour. However, the highly non-linear material response prevents an evaluation by means of established, brittle nature based composite material parameters. Besides the attempt to extract engineering constants at fixed strain values (moduli, strengths), a data-driven approach is deducted enabling the modelling of the stress-strain response at different shear strain rate levels. It is used to generate virtual data to train a machine learning model capable of extracting modelling parameters from experimental results.