2.4. Developed Long-Short Term Memory (LSTM) Model
In the developed study, the data collectors shown in Figure 1, located
in 3 different locations, transferred the humidity and temperature data
of the environment they were in for six months to the server. Since the
locations are used for different purposes, it is seen that the
temperature and humidity values remain within certain ranges. It is
seen that the first location has 0-15° degrees, the second location has
15-25° degrees, and the third location has 23-30° degrees. There are six
different classifications for heating and cooling settings within the
model. It has datasets containing only two classifications in three
locations. Modeling of systems that take such narrow data ranges and
classification with artificial intelligence systems will be limited. The
model learned with data from a single location will not produce correct
results when it encounters data from other classifications. For this
reason, in our study, it is aimed to train the data of different
locations with narrow data ranges on central servers and to transfer
their weights to all locations. Thus, classification will be performed
against a temperature and humidity value encountered for the first time
in the number one location, and the system will continue to operate
without any problems. The data obtained after the data collection
process was stored on the central server. As given in Table 2, six
different order classifications have been determined for the operation
of the heating and cooling system needed according to the ranges of
temperature and humidity values.
Table 2. Temperature ranges and classification groups to be used for
model training