Accident predicting method by data from multiple sensors

The paper describes a method for predicting accidents using real-time readings from multiple sensors. The prediction is made using a set of patterns that describes the changes in the readings of sensors in case of a accident development. The prediction is provided for a given time interval that is sufficient for taking preventive action. The patterns are created as a result of the analysis of sensor readings for the previous time interval. We also describe the transformation of the input data necessary to carry out both the analysis and the predicting.