A Survey of Long-Tail Item Recommendation Methods

Recommender systems represent a critical field of AI technology applications. The core function of a recommender system is to recommend items of interest to users, but if it is only user history-based (purchasing or browsing data), it can only recommend similar products to a user, which makes the user feel fatigued (creating so-called “Information Cocoons”). Besides, transaction data (purchasing or browsing data) in various fields usually follow Pareto distributions. Accordingly, 20% of products are purchased or viewed a greater number of times (short-head items), while the remaining 80% of products are purchased or viewed less frequently (long-tail items). Using the traditional recommendation method, considering only the accuracy of recommendations, the coverage rate is relatively low, and most of the recommended items are short-head items. The long-tail item recommendation method not only considers the recommendation of short-head items but also considers recommending more long-tail items to users, thus improving the coverage and diversity of the recommendation results. Long-tail item recommendation research has become a frontier issue in recommendation systems in recent years. While the current research paper is still scarce, there have been related research achievements in top-level conferences in the field of computers, such as VLDB and IJCAI. Due to the fact that there is no review literature in this field, to allow readers to better understand the research status of the long-tail item recommendation method, this paper summarizes the progress of the research on long-tail item recommendation methods (from clustering-based, which began in 2008, to deep learning-based methods, which began in 2020) and the future directions associated with this research.

How do pandemics end? Two decades of recurrent outbreak risk following the main waves

We analyse the dynamic evolution of disease outbreak risk after the main waves of the 1918-19 “Spanish flu” pandemic in the US and in major cities in the UK, and after the 1890-91 “Russian flu” pandemic in England and Wales. We compile municipal public health records and use national data to model the stochastic process of mortality rates after the main pandemic waves as a sequence of bounded Pareto distributions with an exponentially decaying tail parameter. In all cases, we find elevated mortality risk lasting nearly two decades. An application to COVID-19 under model uncertainty shows that in 90% of model-predicted time series, the annual probability of outbreaks exceeding 500 deaths per million is above 20% for a decade, remaining above 10% for two decades.

How do pandemics end? Two decades of recurrent outbreak risk following the main waves

We analyse the dynamic evolution of disease outbreak risk after the main waves of the 1918-19 “Spanish flu” pandemic in the US and in major cities in the UK, and after the 1890-91 “Russian flu” pandemic in England and Wales. We compile municipal public health records and use national data to model the stochastic process of mortality rates after the main pandemic waves as a sequence of bounded Pareto distributions with an exponentially decaying tail parameter. In all cases, we find elevated mortality risk lasting nearly two decades. An application to COVID-19 under model uncertainty shows that in 80% of model-predicted time series, the annual probability of outbreaks exceeding 500 deaths per million is above 20% for a decade, remaining above 10% for two decades.

Composite Pareto Distributions for Modelling Household Income Distribution in Malaysia

Composite Pareto distributions are flexible as the models allow for data to be described by two distributions: a Pareto distribution for the data above a threshold value and another separate distribution for data below the threshold value. It is noted in some previous literatures that the Paretian tail behaviour can be observed in the distribution of Malaysian household income. In this paper, the composite Pareto models are fitted to the Malaysian household income data of several years. These fitted composite Pareto models are then compared to several univariate models for describing income distribution using pseudo-likelihood based AIC, BIC and Kolmogorov-Smirnov goodness-of-fit test. It is found that the income distributions in Malaysia can be best described by the lognormal-Pareto (II) model as compared to other candidate models.

Matrix variate pareto distributions

Matrix variate generalizations of Pareto distributions are proposed. Several properties of these distributions including cumulative distribution functions, characteristic functions and relationship to matrix variate beta type I and matrix variate type II distributions are studied.

Aspects of Pareto distributions

Different skew models, such as the lognormal and the Pareto functions, have been proposed as suitable descriptions of income distribution. Specific distributions are usually applied in empirical investigations. It is a common opinion that the Pareto curve often provides an adequate description of higher incomes. Recently, double Pareto distributions that obey the power law in both the upper and lower tails have been suggested to reflect a general distribution of personal income. In this study, the literature concerning double Pareto models is presented and the model is applied to Finnish income data. JEL classification numbers: I32. Keywords: Maximum likelihood estimate, Method of moments, Bayesian method, Mean Squared Error, Lognormal, double Pareto, Coefficient of determination, survival function, Geometric Brownian motion.