The cognitive radio is proposed to address the problem of spectrum scarcity in the
current condition where the utilization of the spectrums is imbalance. The users in overutilized
spectrum may access the other licensed spectrum opportunistically without
causing harmful interference to the licensed user. To reduce the interference
harmfulness, the unlicensed users have to detect the presence of the licensed user which
is called spectrum sensing process. There are many problems in spectrum sensing
process. Many works have been done to address the energy consumption problem and
also increase the detection performance. There are several works have been done by
using heuristic approach and also learning approach. Heuristic approaches mostly suffer
from unoptimized results where the learning approaches suffer from long learning time
and also the needed for prediction data. In this work, we propose the combination of
prediction and heuristic approach where the users learn when to use the proper approach
to increase the detection performance and also reduce the energy consumption.