Nowadays, with the rise of location-based services, the personalized sequential POI recommendation has become a pivotal element for enhancing customer experiences. Although many previous POI recommendation models have shown promising results and improvements in this area, several challenges still exist in this field. Firstly, the previous sequential recommenders do not well-utilize the geographical features that are highly affecting the user’s future choices of visits. Furthermore, the self-attention mechanism, which is a popular method used in sequential POI recommendation, has a limitation in treating the input user sequence as an unordered set. Using positional embedding is a typical way to overcome this limitation. However, the use of such embeddings may potentially restrict the model’s ability to learn meaningful patterns in user preferences among POIs. To address these challenges, we propose GeoMixer, a novel MLP-based sequential POI recommender that incorporates travel routing distance to capture geographical features and leverages Multi-layer Perceptron (MLP) architecture to model the spatial and sequential patterns in the sequential POI recommendations. By adopting MLP mixing layers, GeoMixer has the capability of memorizing the chronological order of the input POIs without the positional embedding and can emphasize the important latent features of each POI. The use of the travel routing information improves the model’s ability of capturing spatial patterns during the model learning process. Extensive experiments on real-world datasets show that GeoMixer outperforms state-of-theart methods in various metrics, highlighting the significance of incorporating travel routing distance and leveraging MLP architecture in sequential POI recommendation systems.

We consider the problem of search result diversification for streams of short texts. Diversifying search results in short text streams is more challenging than in the case of long documents, as it is difficult to capture the latent topics of short documents. To capture the changes of topics and the probabilities of documents for a given query at a specific time in a short text stream, we propose a dynamic Dirichlet multinomial mixture topic model, called D2M3, as well as a Gibbs sampling algorithm for the inference. We also propose a streaming diversification algorithm, SDA, that integrates the information captured by D2M3 with our proposed modified version of the PM-2 (Proportionality-based diversification Method -- second version) diversification algorithm. We conduct experiments on a Twitter dataset and find that SDA statistically significantly outperforms state-of-the-art non-streaming retrieval methods, plain streaming retrieval methods, as well as streaming diversification methods that use other dynamic topic models.

The neighbourhood-based Collaborative Filtering is a widely used method in recommender systems. However, the risks of revealing customers' privacy during the process of filtering have attracted noticeable public concern recently. Specifically, kNN attack discloses the target user's sensitive information by creating k fake nearest neighbours by non-sensitive information. Among the current solutions against kNN attack, the probabilistic methods showed a powerful privacy preserving effect. However, the existing probabilistic methods neither guarantee enough prediction accuracy due to the global randomness, nor provide assured security enforcement against kNN attack. To overcome the problems of current probabilistic methods, we propose a novel approach, Probabilistic Partitioned Neighbour Selection, to ensure a required security guarantee while achieving the optimal prediction accuracy against kNN attack. In this paper, we define the sum of k neighbours' similarity as the accuracy metric α, the number of user partitions, across which we select the k neighbours, as the security metric β. Differing from the present methods that globally selected neighbours, our method selects neighbours from each group with exponential differential privacy to decrease the magnitude of noise. Theoretical and experimental analysis show that to achieve the same security guarantee against kNN attack, our approach ensures the optimal prediction accuracy.

We address the problems of privacy-preserving duplicate tuple matching (PPDTM) and privacy-preserving threshold attributes matching (PPTAM) in the scenario of a horizontally partitioned database among N parties, where each party holds a private share of the database's tuples and all tuples have the same set of attributes. In PPDTM, each party determines whether its tuples have any duplicate on other parties' private databases. In PPTAM, each party determines whether all attribute values of each tuple appear at least a threshold number of times in the attribute unions. We propose protocols for the two problems using additive homomorphic cryptosystem based on the subgroup membership assumption, e.g., Paillier's and ElGamal's schemes. By analysis on the total numbers of modular exponentiations, modular multiplications and communication bits, with a reduced computation cost which dominates the total cost, by trading off communication cost, our PPDTM protocol for the semihonest model is superior to the solution derivable from existing techniques in total cost. Our PPTAM protocol is superior in both computation and communication costs. The efficiency improvements are achieved mainly by using random numbers instead of random polynomials as existing techniques for perturbation, without causing successful attacks by polynomial interpolations. We also give detailed constructions on the required zero-knowledge proofs and extend our two protocols to the malicious model, which were previously unknown.

In this paper, we address the problem of cache replacement for transcoding proxy caching. First, an efficient cache replacement algorithm is proposed. Our algorithm considers both the aggregate effect of caching multiple versions of the same multimedia object and cache consistency. Second, a complexity analysis is presented to show the efficiency of our algorithm. Finally, some preliminary simulation experiments are conducted to compare the performance of our algorithm with some existing algorithms. The results show that our algorithm outperforms others in terms of the various performance metrics.

Web caching is an important technique for reducing web traffic, user access latency, and server load and cache replacement plays an important role in the functionality of web caching. In this paper we propose an improved GreedyDual (GD) cache document replacement algorithm, which considers update frequency as a factor in its utility function. We use both trace data and statistical data to simulate our proposed algorithm. The experimental results show that our improved GD algorithm can outperform the existing GD algorithm over the performance metrics considered.

We study the problem of computing classification rule sets from relational databases so that accurate predictions can be made on test data with missing attribute values. Traditional classifiers perform badly when test data are not as complete as the training data because they tailor a training database too much. We introduce the concept of one rule set being more robust than another, that is, able to make more accurate predictions on test data with missing attribute values. We show that the optimal class association rule set is as robust as the complete class association rule set. We then introduce the k-optimal rule set, which provides predictions exactly the same as the optimal class association rule set on test data with up to k missing attribute values. This leads to a hierarchy of k-optimal rule sets in which decreasing size corresponds to decreasing robustness, and they all more robust than a traditional classification rule set. We introduce two methods to find k-optimal rule sets, i.e. an optimal association rule mining approach and a heuristic approximate approach. We show experimentally that a k-optimal rule set generated by the optimal association rule mining approach performs better than that by the heuristic approximate approach and both rule sets perform significantly better than a typical classification rule set (C4.5Rules) on incomplete test data.

Mining transaction databases for association rules usually generates a large number of rules, most of which are unnecessary when used for subsequent prediction. In this paper we define a rule set for a given transaction database that is much smaller than the association rule set but makes the same predictions as the association rule set by the confidence priority. We call this subset the informative rule set. The informative rule set is not constrained to particular target items; and it is smaller than the non-redundant association rule set. We present an algorithm to directly generate the informative rule set, i.e., without generating all frequent itemsets first, and that accesses the database less often than other unconstrained direct methods. We show experimentally that the informative rule set is much smaller than both the association rule set and the non-redundant association rule set, and that it can be generated more efficiently.