Sunday 30 March 2025
Researchers have been working on developing a way to automatically classify products using machine learning algorithms, and their efforts are starting to pay off. In a recent study, scientists used natural language processing techniques to transform product names into numerical vectors, which can then be fed into various classification models.
The team started by collecting a dataset of over 50,000 product names from e-commerce websites. They then manually labeled a subset of these products with specific categories from the ECOICOP international classification system. This labeled data was used to train and test different machine learning algorithms, including logistic regression, decision trees, random forests, and support vector machines.
One of the key challenges in this project was finding an effective way to transform product names into numerical vectors that could be used by the machine learning algorithms. The researchers experimented with several different techniques, including count vectorization, term frequency-inverse document frequency (TF-IDF), Word2Vec, FASTTEXT, and GloVe.
Count vectorization is a simple method that involves counting the frequency of each word in a product name. TF-IDF is a more sophisticated approach that takes into account not only the frequency of each word but also its importance within the overall text. Word2Vec and FASTTEXT are neural network-based methods that use context to predict the meaning of words. GloVe is another neural network-based method that uses global word-word co-occurrence statistics.
The researchers found that different vectorization techniques produced different results, with some performing better than others depending on the specific classification algorithm used. They also discovered that combining multiple vectorization techniques could lead to even better performance.
In addition to experimenting with different vectorization techniques, the team also explored various machine learning algorithms and their hyperparameters. They used a technique called grid search to find the optimal values for each parameter, which involved training and testing many different models.
The results of this study are promising. The researchers found that some machine learning algorithms were able to achieve high levels of accuracy in classifying products, with the best-performing model achieving an accuracy rate of over 99%. They also discovered that certain vectorization techniques were more effective than others depending on the specific classification algorithm used.
This research has important implications for the field of price statistics. By developing a way to automatically classify products using machine learning algorithms, researchers can streamline the process of collecting and analyzing data on prices and product availability. This could lead to more accurate and timely statistical reports, which in turn could inform policy decisions and improve economic outcomes.
Cite this article: “Automated Product Classification Using Machine Learning Algorithms”, The Science Archive, 2025.
Machine Learning, Natural Language Processing, Product Classification, E-Commerce, Dataset, Algorithms, Logistic Regression, Decision Trees, Random Forests, Support Vector Machines
Reference: Bogdan Oancea, “Text classification using machine learning methods” (2025).
