The work described in this article results from a problem proposed by a water utility company in the framework of ESGI 140th, during June 2018. The objective is to evaluate water meters performance using historical data, knowing that, being a mechanical device, the water meters suffer a deterioration with time and use, losing accuracy throughout its cycle of use. We intend to approach a problem capable of identifying anomalies on water consumption pattern. In present work, ARIMA modeling was considered to obtain a predictive model. The results show that in the time series traditional framework revealed significant and adequate in the different estimated models. The in-sample forecast is promising, conducting to adequate measures of performance.

--