A Bayesian Network Model for Interesting Itemsets

Abstract

Mining itemsets that are the most interesting under a statistical model of the underlying data is a commonly used and well-studied technique for exploratory data analysis, with the most recent interestingness models exhibiting state of the art performance. Continuing this highly promising line of work, we propose the first, to the best of our knowledge, generative model over itemsets, in the form of a Bayesian network, and an associated novel measure of interestingness. Our model is able to efficiently infer interesting itemsets directly from the transaction database using structural EM, in which the E-step employs the greedy approximation to weighted set cover. Our approach is theoretically simple, straightforward to implement, trivially parallelizable and retrieves itemsets whose quality is comparable to, if not better than, existing state of the art algorithms as we demonstrate on several real-world datasets.Comment: Supplementary material attached as Ancillary File; in PKDD 2016: European Conference on Machine Learning and Knowledge Discovery in Database

A State-of-the-Art Survey on Deep Learning Theory and Architectures

“…The main idea behind this method is to integrate a probabilistic model (A Gaussian Mixture Model) into the GAN framework that supports likelihood rather than classification [236]. A GAN with Bayesian Network model [237]. Variational Auto encode is a popular deep learning approach, which is trained with Adversarial Variational Bayes (AVB) which helps to establish a principle connection between VAE and GAN [238].…”

Synthetic Dataset Generation with Itemset-Based Generative Models

“…IIM (Interesting Itemset Miner) [4] is an algorithm for mining interesting itemsets from a transactional database. This interestingness of itemsets is defined according to a statistical model, unlike other frequent itemset mining algorithms such as Eclat or FP-Growth which define the interestingness of an itemset based solely on its frequency.…”

“…Algorithm 3: IIM-based generator Algorithm 3 presents our generator implementation using the generative model proposed by [4]. Due to our experimental needs, we generate the same number of transactions as those of the original dataset, yet the generator algorithm is capable of creating synthetic datasets of any given size.…”

“…Among the generative models found in the literature, IGM [5] is the first itemset-based model with a theoretical contribution to the relationship between frequent itemset mining (FIM) and generative models. After that, IIM [4] proposes a generative model over itemsets whose main objective is to find the itemsets that best represent a dataset trying to avoid at the same time redundant ones (which is a typical problem of classic FIM algorithms). LDA [1], on the other hand, was originally proposed to find topics in a set of textual documents.…”

Advanced machine learning algorithms for discrete datasets

“…Top 10 non-singleton patterns selected from the JMLR abstracts dataset to compare pattern interpretability for CDSI (Sec. 2.4), IIM [6] and MTV [7]. .…”