Food Recommendation using Classifier and Modified Apriori Algorithm

Abstract

In today’s world, computer technologies have advanced a lot. One of its greatest gifts to the world is Artificial Intelligence. Natural Language Processing (NLP) and Machine Learning (ML) are two of its subdomains. In this paper, modified versions of two common NLP and ML algorithms have been used to classify food reviews and provide suitable recommendations from them. Currently, reviews can be classified into positive and negative reviews, but it becomes difficult when one review says positive about item A and negative about item B. Moreover, the current Apriori algorithm doesn’t consider the feedbacks from customers (reviews). Modified classifier algorithm and consequently, modified Apriori algorithm has been used to classify each statement part by part and provide recommendations, not just on previous purchases but also using the reviews about above-mentioned purchases. The algorithms can be used for purposes other than food analysis also – wherever purchases and reviews are involved. For e.g., e-commerce companies can use the algorithms to predict and recommend suitable items a user may be interested in.

Analysis of frequent itemset generation based on trie data structure in Apriori algorithm

“…However, this is not the case with Alternative_2. Pruning is done by eliminating [24], [25] the previous long itemset whose frequency is below the minimum support thresholds (infrequent), as seen in Figure 7.…”