Abstract:
This research proposesusingSTackSTorm (ST2) with data cleansing to clean dynamic
data of large sensor networksbefore recording it to the database. There are two cleansing
methods, which are cleansing all data in database before using data and cleansing data at sensors.
Both methods require large amount of processing resources which lead to some problems.
For example, sensor runsout of memory or more energy is needed. Therefore we proposed a
new method of data cleansing, cleansing while gathering the data. With this method, ST2 starts
working when it is triggered. We considered 3 typesof ST triggering, triggered at every event,
triggered once every 10 events and triggered at every 10 second. After ST2 triggered, ran a
prediction algorithm which was a Adams Bashforth method to predict and clean the dirty data.
From the experiment, we found that our method of cleansing while gathering data led to a
decrease in overall processing time. Also, when three types of triggering were compared, we
found that triggering at every event and triggering every 10 seconds gave better performance than
triggering once every 10 events. Both types with better performance gave the same maximum
data transmission rate of 20 data per second without package loss. Also, they provided the same
data cleansing performance in terms of dirty data percentage. In addition, we found that the
number of dirty data did not affect the maximum data transmission rate from sensor networks to
database.