Changing and moving toward online shopping has made it necessary to
customize customers’ needs and provide them more selective options. The
buyers search the products’ features before deciding to purchase items.
The recommender systems facilitate the searching task for customers via
narrowing down the search space within the specific products that align
the customer needs. Clustering, as a typical machine learning approach,
is applied in recommender systems. As an information filtering method, a
recommender system clusters user’s data to indicate the required factors
for more accurate predictions by calculating the similarity between
members of a cluster. In this study, using the Gaussian mixture model
clustering and considering the scores distance and the value of scores
in the Pearson correlation coefficient, a new method is introduced for
predicting scores in machine learning recommender systems. To study the
proposed method’s performance, a Movie Lens data set is evaluated, and
the results are compared to some other recommender systems, including
the Pearson correlation coefficients similarity criteria, K-means, and
fuzzy C-means algorithms. The simulation results indicate that our
method has less error than others by increasing the number of neighbors.
The results also illustrate that when the number of users increases, the
proposed method’s accuracy will increase. The reason is that the
Gaussian mixture clustering chooses similar users and considers the
scores distance in choosing similar neighbors to the active user.
Contextual classification for smart machining based on unsupervised machine learning by Gaussian mixture model
