Search

House price prediction is a popular topic, and research teams are increasingly performing related studies by using deep learning or machine learning models. However, because some studies have not considered comprehensive information that affects house prices, prediction results are not always sufficiently precise. Therefore, we propose an end to end joint self-attention model for house prediction. In this model, we import data on public facilities such as parks, schools, and mass rapid transit stations to represent the availability of amenities, and we use satellite maps to analyze the environment surrounding houses. We adopt attention mechanisms, which are widely used in image, speech, and translation tasks, to identify crucial features that are considered by prospective house buyers. The model can automatically assign weights when given transaction data. Our proposed model differs from self-attention models because it considers the interaction between two different features to learn the complicated relationship between features in order to increase prediction precision. We conduct experiments to demonstrate the performance of the model. Experimental data include actual selling prices in real estate transaction data for the period from 2017 to 2018, public facility data acquired from the Taipei and New Taipei governments, and satellite maps crawled using the Google Maps application programming interface. We utilize these datasets to train our proposed and compare its performance with that of other machine learning-based models such as Extreme Gradient Boosting and Light Gradient Boosted Machine, deep learning, and several attention models. The experimental results indicate that the proposed model achieves a low prediction error and outperforms the other models. To the best of our knowledge, we are the first research to incorporate attention mechanism and STN network to conduct house price prediction.

In today’s more distributed and disorderly network environment, how to organize this information simply and effectively, so that users can quickly obtain potentially valuable data is a common problem in all fields. The commonly used classification systems are based on genetic algorithms and orthogonal decomposition. These two types of systems have high memory usage and low classification accuracy. Aiming at the above problems, a distributed multidimensional big data classification system based on differential equations is designed. The system design is mainly divided into three parts: the first design system overall framework; the second design system hardware, including multidimensional data integration module, central processing module, storage module, result output and display module; third, designing multidimensional big data according to differential equation Classification software main program. The results show that compared with the big data classification system based on genetic algorithm and the big data classification system based on orthogonal decomposition, the classification accuracy of distributed multidimensional big data classification system based on differential equation is improved by 8.75% and 6.75%, and the system memory occupancy rate is improved. Reduce by 35% and 12%.

Aiming at the problems that the traditional method has long response time to communication network congestion monitoring and the detection effect is not ideal, a real-time monitoring method based on cloud computing for communication network blocking is proposed. Firstly, the communication network monitoring point is established, and the communication data collection process is completed by the radio-frequency receiver. On this basis, the real-time traffic calculation of the collected data is performed to determine the existence of abnormal blocking status in the communication network link, and the precise positioning of the blocking point is obtained. The information thus generates an alarm message to obtain a monitoring result. The real-time and accuracy of the monitoring method are analyzed experimentally. It is found that the monitoring method can control the delay time within 0.2 s and the monitoring error rate is low. It can be seen that the monitoring algorithm has high performance.

Summary: The Muscleblind-like protein family (MBNL) plays an important role in regulating the transition between differentiation and pluripotency and in the pathogenesis of myotonic dystrophy type 1 (DM1), a CTG expansion disorder. How different MBNL isoforms contribute to the differentiation and are affected in DM1 has not been investigated. Here, we show that the MBNL1 cytoplasmic, but not nuclear, isoform promotes neurite morphogenesis and reverses the morphological defects caused by expanded CUG RNA. Cytoplasmic MBNL1 is polyubiquitinated by lysine 63 (K63). Reduced cytoplasmic MBNL1 in the DM1 mouse brain is consistent with the reduced extent of K63 ubiquitination. Expanded CUG RNA induced the deubiqutination of cytoplasmic MBNL1, which resulted in nuclear translocation and morphological impairment that could be ameliorated by inhibiting K63-linked polyubiquitin chain degradation. Our results suggest that K63-linked ubiquitination of MBNL1 is required for its cytoplasmic localization and that deubiquitination of cytoplasmic MBNL1 is pathogenic in the DM1 brain. : Wang et al. find that MBNL1 ubiquitination is required for cytoplasmic localization and promotion of neurite outgrowth. In myotonic dystrophy, expanded CUG repeat RNA leads to MBNL1 deubiquitination, resulting in nuclear-translocation-associated morphological defects that can be rescued by preventing degradation of lysine 63-linked polyubiquitin chains or enhancing MBNL1 ubiquitination. Keywords: MBNL1, polyubiquitination, myotonic dystrophy, nucleocytoplasmic localization, neurite outgrowth

With rapid developments in deep learning and financial technology, a customized robo-advisory service based on novel artificial intelligence techniques has been widely adopted to realize financial inclusion. This study proposes a novel robo-advisor system that integrates trend prediction, portfolio management, and a recommendation mechanism. A gated neural network structure combining three multiobjective RankNet kernels could rank target financial products and recommend the top-n securities to investors. The gated neural network learns to choose or weigh each RankNet for incorporating the most important partial network inputs, such as earnings per share, market index, and hidden information from the time series. Experimental results indicate that the recommendation results of our proposed robo-advisor based on a gated neural network and multiobjective RankNets can outperform existing models.

Background In this study, we investigated the association of time-varying serum albumin levels with mortality over a 5-year period in one cohort of patients undergoing long-term peritoneal dialysis (PD) therapy. Methods The participants in this study enrolled 302 patients who underwent long-term PD at a single PD center in Taiwan. We reviewed medical records from 2011 to 2015 retrospectively. Time-averaged albumin level and serum albumin reach rate (defined as the percentage of serum albumin measurements that reached ≥3.5 g/dL) were applied as the predictor variables in the first 2 years (2011–2012). All-cause mortality was used as the outcome variable in the subsequent 3 years (2013–2015). Hazard function of all-cause mortality in the study participants was examined by using Cox proportional hazard regression models . Results Patients with different albumin reach rates (75–

Myotonic dystrophy type 1 (DM1) is caused by an expansion of CTG repeats in the 3' untranslated region (UTR) of the dystrophia myotonia protein kinase (DMPK) gene. Cognitive impairment associated with structural change in the brain is prevalent in DM1. How this histopathological abnormality during disease progression develops remains elusive. Nuclear accumulation of mutant DMPK mRNA containing expanded CUG RNA disrupting the cytoplasmic and nuclear activities of muscleblind-like (MBNL) protein has been implicated in DM1 neural pathogenesis. The association between MBNL dysfunction and morphological changes has not been investigated. We generated a mouse model for postnatal expression of expanded CUG RNA in the brain that recapitulates the features of the DM1 brain, including the formation of nuclear RNA and MBNL foci, learning disability, brain atrophy and misregulated alternative splicing. Characterization of the pathological abnormalities by a time-course study revealed that hippocampus-related learning and synaptic potentiation were impaired before structural changes in the brain, followed by brain atrophy associated with progressive reduction of axon and dendrite integrity. Moreover, cytoplasmic MBNL1 distribution on dendrites decreased before dendrite degeneration, whereas reduced MBNL2 expression and altered MBNL-regulated alternative splicing was evident after degeneration. These results suggest that the expression of expanded CUG RNA in the DM1 brain results in neurodegenerative processes, with reduced cytoplasmic MBNL1 as an early event response to expanded CUG RNA.