Sales forecasting plays a critical role in optimizing supply chains, reducing costs, and enhancing customer satisfaction within enterprises. The presence of demand volatility, seasonality, and non-linear relationships with products pose higher requirements for forecasting models. Although deep learning models have demonstrated promising performance in addressing these challenges, the majority of research has primarily focused on modeling historical sales variations and product interactions in the time domain or solely considered periodic representations in the frequency domain within the models. We propose a novel deep learning model TSFRN based on the triple-links residual networks to integrate the time and frequency domain information in this paper, which consists of a forecast link, spatial-frequency forecast link, and a backward link. Specifically, in each block of the network, we first use a recurrent neural network to obtain a time domain prediction of historical sales as the forecast link. Next, we apply the Fast Fourier Transform (FFT) to obtain the frequency domain representation of sales data. Subsequently, a spatial-frequency domain attention mechanism is proposed in this study to capture spatial interaction patterns within the frequency domain of the sales data. Finally, we obtain the prediction based on the frequency domain through an inverse fast Fourier transform as the spatial-frequency forecast link. For the backward link, we minus the forecast link by the input of the current block. Overall, our proposed framework integrates both time domain and frequency domain information to model the complex interrelationships among products. By taking into account both model construction and practical applications, our framework provides a more accurate, effective, and general approach to sales forecasting. Experimental validation on publicly available datasets demonstrates the effectiveness of our proposed model, which outperforms existing methods in predicting product sales with improved accuracy.

Hierarchical time series forecasting plays a vital role in practical applications. To guarantee the accuracy and coherence simultaneously for the prediction, most of the classical methods tackle this problem with two-stage models, which may not obtain optimal predictions due to the complex computational process. Although the majority of end-to-end deep learning models were proposed to avoid the shortage of traditional methods, most existing works stack the shallow neural network directly which caused degradation issues possibly and decreased the prediction accuracy. In this paper, we draw on the idea of deep residual networks (ResNet) and the top-down method to form an end-to-end hierarchical decomposition residual network (HDResNet), which could decrease the computation cost of prediction and improve the forecasting outcome. For each block of our model, we firstly employ the Long Short-Term Memory (LSTM) layer and linear layer to obtain two decompositions, the residual forecasting and proportion forecasting. Then, a functional layer which is based on the top-down approach is proposed, to update the residual forecasting of children nodes and decrease the noise transmission. Moreover, considering the stability at different levels, a weighted loss function is proposed based on the coefficient of variation for each level. Through three public datasets, we prove that our model has a superior performance to the existing hierarchical forecasting models.

Constructing a quantitative factor investment strategy based on hundreds of candidate factors is a critical challenge. Existing linear models do not account for nonlinearities and variable interactions, while complex machine learning models are easily overfitting. In this paper, motivated by the portfolio sorts methods in empirical asset pricing, we propose an alternative approach called grouping-based AdaBoost by adapting the existing AdaBoost. It introduces the experience of the financial field into the algorithm design to improve the performance and generalization of machine learning-based factor investing strategies. The proposed method restricts the factor to only predict the common part of the returns of the same groups and allows the potential nonlinear relationship between a factor and the return. Moreover, to enhance the model's ability to use factors with high correlation, we extend the single-grouping AdaBoost in a multi-grouping way. Experiments on the Chinese A-share market demonstrate the effectiveness of our approach in both stock performance classification and portfolio selection and provide intuitive evidence for the generalization of the proposed method.

This work takes the lead to study the aspect-level sentiment classificationin the domain adaptation scenario. Given a document of any domains, the model needs to figure out the sentiments with respect to fine-grained aspects in the documents. Two main challenges exist in this problem. One is to build a robust document modeling across domains; the other is to mine the domain-specific aspects and make use of the sentiment lexicon. In this paper, we propose a novel approach Neural Attentive model for cross-domain Aspect-level sentiment CLassification (NAACL), which leverages the benefits of the supervised deep neural network as well as the unsupervised probabilistic generative model to strengthen the representation learning. NAACL jointly learns two tasks: (i) a domain classifier, working on documents in both the source and target domains to recognize the domain information of input texts and transfer knowledge from the source domain to the target domain. In particular, a weakly supervised Latent Dirichlet Allocation model (wsLDA) is proposed to learn the domain-specificaspectandsentiment lexiconrepresentations that are then used to calculate the aspect/lexicon-aware document representations via a multi-view attention mechanism; (ii) an aspect-level sentiment classifier, sharing the document modeling with the domain classifier. It makes use of the domain classification results and the aspect/sentiment-aware document representations to classify the aspect-level sentiment of the document in domain adaptation scenario. NAACL is evaluated on both English and Chinese datasets with the out-of-domain as well as in-domain setups. Quantitatively, the experiments demonstrate that NAACL has robust superiority over the compared methods in terms of classification accuracy and F1 score. The qualitative evaluation also shows that the proposed model is capable of reasonably paying attention to those words that are important to judge the sentiment polarity of the input text given an aspect.

Open-domain dialog generation, which is a crucial component of artificial intelligence, is an essential and challenging problem. In this article, we present a personalized dialog system, which leverages the advantages of multitask learning and reinforcement learning for personalized dialogue generation (MRPDG). Specifically, MRPDG consists of two subtasks: 1) an author profiling module that recognizes user characteristics from the input sentence (auxiliary task) and 2) a personalized dialog generation system that generates informative, grammatical, and coherent responses with reinforcement learning algorithms (primary task). Three kinds of rewards are proposed to generate high-quality conversations. We investigate the effectiveness of three widely used reinforcement learning methods [i.e., Q-learning, policy gradient, and actor-critic (AC) algorithm] in a personalized dialog generation system and demonstrate that the AC algorithm achieves the best results on the underlying framework. Comprehensive experiments are conducted to evaluate the performance of the proposed model on two real-life data sets. Experimental results illustrate that MRPDG is able to produce high-quality personalized dialogs for users with different characteristics. Quantitatively, the proposed model can achieve better performance than the compared methods across different evaluation metrics, such as the human evaluation, BiLingual Evaluation Understudy (BLEU), and perplexity.

As an increasing number of investors post their opinions or show their decisions on public platforms, a critical challenge is to make trading decisions by considering opinions from online investors. In this paper, by taking the real-world data from Stocktwits as an example, we develop FollowAKOInvestor, a systematic two-step framework to utilize sentiments from various kinds of investors to forecast stocks. First, FollowAKOInvestor divides investors into various groups according to their expertise levels and extract sentiments from different investor groups as features. Then, it uses machine learning techniques to make trading decisions by learning a prediction function to appropriately combine sentiments from different kinds of investors. The intuition of FollowAKOInvestor is that sentiments extracted from all kinds of investors (including experts and non-experts) can help us invest in stocks. Extensive data analysis and experiments show that FollowAKOInvestor generates high-yield portfolios.

The recent artificial intelligence studies have witnessed great interest in abstractive text summarization. Although remarkable progress has been made by deep neural network based methods, generating plausible and high-quality abstractive summaries remains a challenging task. The human-like reading strategy is rarely explored in abstractive text summarization, which however is able to improve the effectiveness of the summarization by considering the process of reading comprehension and logical thinking. Motivated by the humanlike reading strategy that follows a hierarchical routine, we propose a novel Hybrid learning model for Abstractive Text Summarization (HATS). The model consists of three major components, a knowledge-based attention network, a multitask encoder-decoder network, and a generative adversarial network, which are consistent with the different stages of the human-like reading strategy. To verify the effectiveness of HATS, we conduct extensive experiments on two real-life datasets, CNN/Daily Mail and Gigaword datasets. The experimental results demonstrate that HATS achieves impressive results on both datasets.

It is very useful to endow machines with the ability to understand users' personalized preferences. In this paper, we propose a novel methodology for discovering investors' personalized preferences in stock markets. Our work is able to estimate investors' personalized preferences for each stock and thus helpful for realizing investment recommendation, for instance through recommending real-time news or others' opinions on stocks preferred by the target user. Compared to conventional approaches, our method effectively incorporates stock movements for estimating investors' preference. By capturing stock-movement patterns influencing users' preferences, our method can find users with a similar investment philosophy and then increase the effect of preference prediction. An experimental evaluation with two real-world datasets demonstrates the effectiveness of our approach.

Recommending social activities , such as watching movies or having dinner, is a common function found in social networks or e-commerce sites. Besides certain websites which manage activity-related locations (e.g., foursquare.com), many items on product sale platforms (e.g., groupon.com) can naturally be mapped to social activities. For example, movie tickets can be thought of as activity items , which can be mapped as a social activity of “watch a movie.” Traditional recommender systems estimate the degree of interest for a target user on candidate items (or activities), and accordingly, recommend the top- k activity items to the user. However, these systems ignore an important social characteristic of recommended activities: people usually tend to participate in those activities with friends. This article considers this fact for improving the effectiveness of recommendation in two directions. First, we study the problem of activity-partner recommendation ; i.e., for each recommended activity item, find a suitable partner for the user. This (i) saves the user’s time for finding activity partners, (ii) increases the likelihood that the activity item will be selected by the user, and (iii) improves the effectiveness of recommender systems to users overall and enkindles their social enthusiasm. Our partner recommender is built upon the users’ historical attendance preferences, their social context, and geographic information. Moreover, we explore how to leverage the partner recommendation to help improve the effectiveness of recommending activities to users. Assuming that users tend to select the activities for which they can find suitable partners, we propose a partner-aware activity recommendation model, which integrates this hypothesis into conventional recommendation approaches. Finally, the recommended items not only match users’ interests, but also have high chances to be selected by the users, because the users can find suitable partners to attend the corresponding activities together. We conduct experiments on real data to evaluate the effectiveness of activity-partner recommendation and partner-aware activity recommendation. The results verify that (i) suggesting partners greatly improves the likelihood that a recommended activity item is to be selected by the target user and (ii) considering the existence of suitable partners in the ranking of recommended items improves the accuracy of recommendation significantly.

We present an attention-based bidirectional LSTM approach to improve the target-dependent sentiment classification. Our method learns the alignment between the target entities and the most distinguishing features. We conduct extensive experiments on a real-life dataset. The experimental results show that our model achieves state-of-the-art results.

Recently, researchers found a new type of attacks, called time synchronization attack (TS attack), in cyber-physical systems. Instead of modifying the measurements from the system, this attack only changes the time stamps of the measurements. Studies show that these attacks are realistic and practical. However, existing detection techniques, e.g. bad data detection (BDD) and machine learning methods, may not be able to catch these attacks. In this paper, we develop a "first difference aware" machine learning (FDML) classifier to detect this attack. The key concept behind our classifier is to use the feature of "first difference", borrowed from economics and statistics. Simulations on IEEE 14-bus system with real data from NYISO have shown that our FDML classifier can effectively detect both TS attacks and other cyber attacks.

Investor social media, such as StockTwist, are gaining increasing popularity. These sites allow users to post their investing opinions and suggestions in the form of microblogs. Given the growth of the posted data, a significant and challenging research problem is how to utilize the personal wisdom and different viewpoints in these opinions to help investment. Previous work aggregates sentiments related to stocks and generates buy or hold recommendations for stocks obtaining favorable votes while suggesting sell or short actions for stocks with negative votes. However, considering the fact that there always exist unreasonable or misleading posts, sentiment aggregation should be improved to be robust to noise. In this paper, we improve investment recommendation by modeling and using the quality of each investment opinion. To model the quality of an opinion, we use multiple categories of features generated from the author information, opinion content and the characteristics of stocks to which the opinion refers. Then, we discuss how to perform investment recommendation (including opinion recommendation and portfolio recommendation) with predicted qualities of investor opinions. Experimental results on real datasets demonstrate effectiveness of our work in recommending high-quality opinions and generating profitable investment decisions.

Discovering the author's interest over time from documents has important applications in recommendation systems, authorship identification and opinion extraction. In this paper, we propose an interest drift model (IDM), which monitors the evolution of author interests in time-stamped documents. The model further uses the discovered author interest information to help finding better topics. Unlike traditional topic models, our model is sensitive to the ordering of words, thus it extracts more information from the semantic meaning of the context. The experiment results show that the IDM model learns better topics than state-of-the-art topic models.

A large-scale training corpus consisting of microblogs belonging to a desired category is important for high-accuracy microblog retrieval. Obtaining such a large-scale microblgging corpus manually is very time and labor-consuming. Therefore, some models for the automatic retrieval of microblogs froman exterior corpus have been proposed. However, these approaches may fail in considering microblog-specific features. To alleviate this issue, we propose a methodology that constructs a simulated microblogging corpus rather than directly building a model from the exterior corpus. The performance of our model is better since the microblog-special knowledge of the microblogging corpus is used in the end by the retrieval model. Experimental results on real-world microblogs demonstrate the superiority of our technique compared to the previous approaches.

Users commonly use Web 2.0 platforms to post their opinions and their predictions about future events (e.g., the movement of astock). Therefore, opinion mining can be used as a tool for predicting future events. Previous work on opinion mining extracts from the text only the polarity of opinions as sentiment indicators. We observe that a typical opinion post also contains temporal references which can improve prediction. This short paper presents our preliminary work on extracting reference time tagsand integrating them into an opinion mining model, in order to improvethe accuracy of future event prediction. We conduct anexperimental evaluation using a collection of microblogs posted by investors to demonstrate the effectiveness of our approach.

Topic modeling of textual corpora is an important and challenging problem. In most previous work, the “bag-of-words” assumption is usually made which ignores the ordering of words. This assumption simplifies the computation, but it unrealistically loses the ordering information and the semantic of words in the context. In this paper, we present a Gaussian Mixture Neural Topic Model (GMNTM) which incorporates both the ordering of words and the semantic meaning of sentences into topic modeling. Specifically, we represent each topic as a cluster of multi-dimensional vectors and embed the corpus into a collection of vectors generated by the Gaussian mixture model. Each word is affected not only by its topic, but also by the embedding vector of its surrounding words and the context. The Gaussian mixture components and the topic of documents, sentences and words can be learnt jointly. Extensive experiments show that our model can learn better topics and more accurate word distributions for each topic. Quantitatively, comparing to state-of-the-art topic modeling approaches, GMNTM obtains significantly better performance in terms of perplexity, retrieval accuracy and classification accuracy.

Min Yang, Wenting Tu, Ziyu Lu, Wenpeng Yin, Kam-Pui Chow. Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2015.

In transfer learning scenarios, previous discriminative dimensionality reduction methods tend to perform poorly owing to the difference between source and target distributions. In such cases, it is unsuitable to only consider discrimination in the low-dimensional source latent space since this would generalize badly to target domains. In this paper, we propose a new dimensionality reduction method for transfer learning scenarios, which is called transferable discriminative dimensionality reduction (TDDR). By resolving an objective function that encourages the separation of the domain-merged data and penalizes the distance between source and target distributions, we can find a low-dimensional latent space which guarantees not only the discrimination of projected samples, but also the transferability to enable later classification or regression models constructed in the source domain to generalize well to the target domain. In the experiments, we firstly analyze the perspective of transfer learning in brain-computer interface (BCI) research and then test TDDR on two real datasets from BCI applications. The experimental results show that the TDDR method can learn a low-dimensional latent feature space where the source models can perform well in the target domain.

Extreme energy ratio (EER) is a recently proposed feature extractor to learn spatial filters for electroencephalogram (EEG) signal classification. It is theoretically equivalent and computationally superior to the common spatial patterns (CSP) method which is a widely used technique in brain-computer interfaces (BCIs). However, EER may seriously overfit on small training sets due to the presence of large noise. Moreover, it is a totally supervised method that cannot take advantage of unlabeled data. To overcome these limitations, we propose a regularization constraint utilizing local temporal information of unlabeled trails. It can encourage the temporal smoothness of source signals discovered, and thus alleviate their tendency to overfit. By combining this regularization trick with the EER method, we present a semi-supervised feature extractor termed semi-supervised extreme energy ratio (SEER). After solving two eigenvalue decomposition problems, SEER recovers latent source signals that not only have discriminative energy features but also preserve the local temporal structure of test trails. Compared to the features found by EER, the energy features of these source signals have a stronger generalization ability, as shown by the experimental results. As a nonlinear extension of SEER, we further present the kernel SEER and provide the derivation of its solutions.