With the explosion of the amount of medical data, data-driven medical analyses are receiving increasing attention. Proper representation of medical activities is crucial for such analyses. However, most existing representations are designed without considering the temporality and numerical sensitivity of medical data, which limits the performance and interpretability of the analysis tasks. This paper presents a representation learning approach for medical activities that is enhanced by topical modeling for inpatient data. The approach leverages the temporal relations between activities and the topic assignment to construct a multilayer perceptron model. Evaluations using large real data sets demonstrate that this approach significantly improves three typical medical analysis tasks, while providing medical interpretations.

Social networking feeds often include much spam that includes marketing, recruitment or short articles without real content which negatively affect the user interest. The spam also seriously affects academic research and business applications. This paper presents an algorithm based on the pSVM-kNN model for filtering Chinese microblogging text noise to reduce the spam. This method combines the SVM and kNN algorithms. The kNN algorithm iteratively finds the optimal solution of the classification hyperplane in the local scope on the SVM computing hyperplane. Penalty costs and proportional weights are introduced into the SVM and kNN stages to improve the noise filtering and reduce misclassification. Tests on various size of real Sina Weibo datasets demonstrate that the precision and recall of this algorithm are significantly better than other methods with a remarkable improvement of the F-measure.

Using massive location data to analyze user behavior and tap the potential value of customers is getting much attention. Mature solutions have been studied for outdoor space, while the accuracy and robustness of WiFi positioning data are main challenges for indoor space. The method of extracting the semantic trajectory for indoor space is studied. It can reduce error, compress original position data and enhance the expression capability of trajectory for indoor spatio-temporal data mining. Based on four modules including indoor space modeling, data cleansing, event extraction and semantic enhancement, a framework for calculating indoor semantic trajectory is proposed. Experiments are conducted on real and simulated datasets, and the results show that this method can accurately and effectively extract trajectory data containing semantic information from indoor positioning data with incompleteness and errors, which can be used for high-level analytical applications.

Knowledge graph data has large volumes, rich content, diverse types, and lacks a unified model description. Pattern information needs to be extracted from knowledge graphs to improve the quality of knowledge graph retrieval and mining. This paper presents a knowledge graph summarization pattern and quality metrics. This method is used in an algorithm for mining Top-k summarization patterns (Top-k SPM) formulated as a submodular function optimization problem. Then, a Pregel based parallel algorithm is used to validate the algorithm and measure the qualities of summarization patterns. Two efficient greedy algorithms are also presented to solve the Top-k SPM. The efficiency and effectiveness of the method is then verified on real knowledge graph datasets. The tests show that the method outperforms the existing methods in terms of coverage and algorithm execution time.

Most existing methods for differential privacy streaming data statistical release based on tree structures cannot take advantage of the special probability distributions in statistical queries. This study presents an algorithm that further boosts the accuracy of the released streaming data for differential privacy streaming data publication with non-uniform private budgets using sliding windows. After constructing the differential privacy range tree for the streaming data within the sliding window, the algorithm calculates the coverage probability of the tree nodes according to the probability distribution of the statistical queries and then adds non-uniform noise to the tree nodes based on the adjusted privacy budget of the tree nodes and a tree structure parameter. Finally, several real-time adjustment policies are designed to ensure that the node values on any path from the root to the leaves satisfy a consistency constraint. Tests show that the algorithm guarantees better statistical query accuracy and has higher algorithm efficiency than traditional algorithms.

Fractures of the lower extremity long bones are common in motor vehicle crashes and human finite-element models are very effective in studies of lower extremity injury risks and injury mechanisms. The subject characteristics (sex, age, stature and weight) significantly affect the bone geometry and the cortical bone thickness distribution. This study used 95 Chinese clinical CTs to develop femur and tibia parametric models using a male mid-size template model through morphing, fitting, and statistical analyses. The average absolute errors in the predicted external surface geometry models were less than 3 mm and that the average absolute errors in the predicted thickness models were less than 0.6 mm. Bone length was the most significant predictor for the bone geometry models. Age and BMI were both significant in predicting the femoral cortex thickness distribution while only age was significant for predicting the tibia's thickness distribution.

Corrugated sleeves are cylindrical shells with a special bulge in the central region, which are also called corrugated sleeves in automotive engineering. Both ends of the corrugated sleeve are straight. Corrugated sleeves have a distinct flat region in their axial force-compressive displacement curves. Corrugated sleeves with this axial load characteristic are used to preload bearings and absorb energy. Therefore, the axial characteristics are quite important in many designs. The axial compression of a corrugated sleeve was simulated here using the finite element method in ABAQUS with the results verified against tests. The axial force-displacement curves were predicted for various parameters to analyze the influence of the geometric parameters on the axial load characteristics. The FEA results were then used with multi-objective optimization to minimize the maximum equivalent plastic strain and maximize the platform width using the Non-dominated Sorting Genetic Algorithm Ⅱ. This method can be used to optimize the designs of corrugated sleeves for various conditions.

The platooning of automated vehicles can greatly improve the fuel efficiency of road transportation. Current studies will reduce fuel consumption by using very short car-following distances to reduce the aerodynamic drag on the following vehicles, which increases rear-end collision risks. This paper presents a periodic control method, pulse and glide (PnG) operation, for heterogeneous vehicle platoons to minimize the overall fuel consumption. The switching control method controls each vehicle to maintain the given the inter-vehicle distance while reducing the fuel consumption. Simulations show that this controller can maintain the inter-vehicle distance within a bounded range and reduce fuel consumption by up to 23.3% compared to linear quadratic (LQ) controllers.

Micro sensing systems will be widely used in the future because they are small, multifunctional and self-powered. However, current designs do not balance their performance with key constraints which limits their further development. This paper presents a general design method for micro sensing systems based on multidisciplinary design optimization. The method facilitates optimization of micro sensing systems by analyzing the parameters in various subspaces to optimize the design. Vehicle detection micro-systems are used as an example with various optimization design targets for the prototypes. The test results coincide well with the design predictions and meet the design requirements. The method accuracy shows that it balances performance requirements with systems constraints during the design process to optimize the system.

The accuracies of several outdoor environment simulation software packages were evaluated for urban radiation field using the city microclimate simulation tool ENVI-met, a human body thermal radiation evaluation software (radiation on the human body, RayMan) and a climate design tool (solar and longwave environmental irradiance geometry, SOLWEIG). First, the radiation distribution predictions in an urban environment were compared. Then, the effects of the surface and building enclosure characteristics were evaluated for the radiation field for long and short wave radiation and direct and reflected radiation. RayMan was found to very sensitive and fast, but output fewer radiation parameters. ENVI-met calculated the most radiation parameters and included the heat transfer from the long and short wave radiation between building surfaces. SOLWEIG can be used for large outdoor space simulations and calculated many radiation parameters, but had significant errors. Thus, these software packages can be used to analyze various influences of enclosures on direct and scattered radiation.