Prediction model of hourly water consumption in water purification plant through categorical approach

In a water purification plant a precise prediction of hourly water consumption is needed for supplying water stably to consumers and to operate the plant efficiently. Hourly water consumption is regarded as time series data with a period of 24 hours. Hourly water consumption data per day, which we call a waveform, reflect the style of our living. The waveforms in weekdays are influenced by fluctuation factors such as the day of the week, weather and temperature but they resemble each other. On the other hand, the data on national holidays or consecutive holidays are not similar to that of usual days and it is more difficult to predict them precisely than usual days. The objectives of our research are to precisely predict the hourly water consumption for the next day especially on such unusual days. We analyzed and categorized hourly water consumption data gathered in a water purification plant in a metropolitan area in Japan for several years with the data mining concept in mind and tried to construct a precise prediction model through the year