Passive biological cilia function as sensory organelles in various animal cells and microorganisms. Employing computational analysis, we seek to understand the hydrodynamic mechanisms that govern the sensing abilities of cilia arrays exposed to particulate flow. The outcomes of simulations depicting cilia and particle interactions can serve as valuable data for machine learning algorithms, aiding in the construction of models capable of predicting particle properties. Thus the aims of this research work are twofold: to efficiently generate enough data for training and to develop a machine learning algorithm for accurate predictions. Large scale data generation necessitates the development of a computationally inexpensive and accurate (two-dimensional) model for cilia and particles. In this study, we propose such a model by adapting an existing framework based on Kirchoff rod theory to represent cilia. The particle is modeled as a neo-Hookean massless solid and solved using the Finite Element Method. Coupling of the cilia and particle with the fluid is accomplished through the Immersed Boundary Method. We conduct numerous simulations involving particles of diverse shapes and sizes traversing an array of cilia within a channel. The movement of the particles is induced by the oscillating top wall of the channel. A machine learning model, comprising a Long Short-Term Memory Network coupled with a Regression Layer, is trained using the generated data, which includes cilia base deflections. Using data from unseen simulations we show that the trained model is indeed capable of predicting the size and aspect ratio of the particle within 5% accuracy.