Motivated by our work in pediatric research, a multilinear normal mixture model is proposed to cluster five way data. Five way data consists of a sample of mode four multidimensional arrays, one per subject. The model has interpretable parameters for the mean multilinear array and the covariance matrices for each mode of the array. The model parameters are estimated using an EM algorithm. Our model is implemented in Julia, a dynamic and high performance numerical computing language. The five way data was generated from accelerometers worn by a cohort of school aged children.