Page 31 - 2023

Our research work proposes a new multi-source sentiment analysis model that includes
various stages to perform the task of domain adaptation. An enhanced cross entropy
measure is exploited to check the similarity between statements of different domains,
which finds the important common features in the target domain and assigns the class
label. The remaining features are then given to the proposed classifier that predicts the
polarity of the target domain in a precise way. These newly found features from the target
domain help the classifier to get trained on some of the target domain features, which
boosts the accuracy value. For classification purposes, Neural Network (NN) is exploited.
Particularly, the weights of NN are tuned in an optimal manner using the Improved Grey
Wolf Optimization (IGWO) algorithm, which is the enhanced version of the GWO
algorithm. GWO algorithm is modified to remove the problem of local optima and speed
up the process of finding the optimal value. The comparison of IGWO and other variants
of GWO is also performed on various optimization-based functions to prove the
superiority of the IGWO in obtaining the global optima with fewer iterations. We have
considered various important performance measures like accuracy, precision, recall, F-
measure, and negative predictive value (NPV) to evaluate the classification result. The
comparison of our classification model with the traditional approaches on the Amazon
review dataset has shown the superiority of the proposed method. Our classification
model classifies the data with improved accuracy of around 28% to 6% compared to
traditional approaches.
Key words: Multi-domain sentiment analysis, Domain adaptation, Enhanced cross
entropy, Improved GWO algorithm

06