Application of Machine Learning in Recommendation Algorithms

Abstract

With the development of technology, the importance of recommendation algorithms in online transactions is gradually increasing. The development of machine learning is essential to improve the performance of recommendation algorithms. Therefore, a suitable model is urgently needed for recommendation algorithms. In order to evaluate the performance of classical machine learning algorithms against newer machine learning algorithms, four models were selected, namely Decision Tree model, Random Forest model, Gradient Boosting Decision Tree model and XGBoost model. They were subjected to regression analysis on the same dataset, and the final scores were derived from the four parameters r2_score, MSE, MAE, and expected variance. The article selects a set of recommendation dataset from Amazon platform and processes the noisy data in the dataset. The data is processed by three methods, namely direct deletion, character digitisation and natural language processing, to obtain a valid dataset of 19, 676 items, and divided into training and prediction sets according to 7:3. Next, the models included in the sklearn library in PYTHON were used for regression analysis to obtain the performance scores of each model. The Decision Tree model had the lowest score of 0.70584 and the Gradient Boosting Decision Tree model had the highest score of 0.85281. In addition, in general, the Decision Tree model scored much worse than the other three models, while the other three models scored more similarly.

Application test of PCA based improved K-means clustering algorithm in analyzing NGO assistance needs in less developed countries

“…There are many forms and standards of machine learning. It is mainly divided into three categories, namely supervised learning, unsupervised learning and reinforcement learning [2]. K-Means clustering is one of the most common clustering algorithms in the field of machine learning.…”

“…However, due to the shortcomings of K-means itself, the dimensional attributes of the data set must be converted into numerical types through arithmetic means to measure the distance. Different random selections will have a certain degree of influence on the final clustering results, and eventually lead to excessive decision-making bias [2] [3]. Especially high noise points, multidimensional and nonlinear social big data have many error values that affect clustering [4].…”