Since manual sorting of apples is a huge project in our daily life, an apple image classification approach based on multi-perspective multi-region feature fusion was proposed. First of all, five classes of apples, containing 329 in total, were collected. For each apple, five images from five different perspectives were obtained:top, bottom, side1, side2 and side3. From each image, several (one to nine) small image regions were cut. Secondly, each region block was represented by color histogram vector, and the histogram vectors of region blocks were fused together end to end to generate a representation of the image. Finally, 12 classifiers were used to classify 329 samples. The experimental results show that the multi-perspective multi-region based method significantly outperforms single-perspective single-region based method, and the more the number of perspectives/regions, the better the result. In particular, classification performance reaches 97.87% by PLS (Partial Least Squares) even better than deep learning when using nine regions for each image cropped at five angles. The method is easy but efficient, whose computation complexity is 4 n, where n is the total number of blocks in image cropping area. Thus, it can be applied to mobile applications and applied to more fruit and plant image classification.

To solve the problem that the existing methods based on the fixed-order Markov models cannot make full use of the structural features involved in the subsequences of different orders, a new Bayesian method based on the multi-order Markov model was proposed for symbolic sequences classification. First, a Conditional Probability Distribution (CPD) model was built based on the multi-order Markov model. Second, a suffix tree for n-order subsequences with efficient suffix-tables and its efficient construction algorithm were proposed, where the algorithm could be used to learn the multi-order CPD models by scanning once the sequence set. A Bayesian classifier was finally proposed for the classification task. The training algorithm was designed to learn the order-weights for the models of different orders based on the Maximum Likelihood (ML) method, while the classification algorithm was defined to carry out the Bayesian prediction using the weighted conditional probabilities of each order. A series of experiments were conducted on real-world sequence sets from three domains and the results demonstrate that the new classifier is insensitive to the predefined order change of the model. Compared with the existing methods such as the support vector machine using the fixed-order model, the proposed method can achieve more than 40% improvement on both gene sequences and speech sequences in terms of classification accuracy, yielding reference values for the optimal order of a Markov model on symbolic sequences.

In view of the important nodes (including overlapping nodes, central nodes and outlier nodes) in overlapping community and the inherent overlapping community structure discovery problem, a new symmetric nonnegative matrix factorization algorithm was proposed. First, the sum of the error approximation and the asymmetric penalty term was used as the objective function. Then the algorithm was derived by using the principle of gradient update and the nonnegative constraint conditions. Simulation experiments were carried out on five real networks. The results show that the proposed algorithm can find the important nodes of the actual networks and their inherent community structures. The average conductance and the algorithm's execution time of the community discovery results are better than those of Community Detection with Nonnegative Matrix Factorization (CDNMF) method;the weighted average of the accuracy and recall rate's harmonic mean value shows that the proposed method is more suitable for the large databases.

A Quadratic path planning algorithm based on sliding window and Ant Colony Optimization (QACO) algorithm was put forward on the issue of weak planning ability of Ant Colony Optimization (ACO) algorithm in complex environments. The feedback strategy of the ACO based on Feedback Strategy (ACOFS) algorithm was improved, and the feedback times were reduced through the decrease of pheromone along feedback path. In the first path planning, the improved ACO algorithm was applied to make a global path planning for the grid environment. In the second path planning, the sliding windows slid along the global path. Local path in sliding windows was planned with ACO algorithm. Then the global path could be optimized by local path until target location was contained in the sliding window. The simulation experiments show that, the average planning time of QACO algorithm respectively reduces by 26.21%, 52.03% and the average length of path reduces by 47.82%, 42.28% compared with the ACO and QACO algorithms. So the QACO algorithm has a relatively strong path planning ability in complex environments.

A new soft subspace clustering algorithm was proposed to address the optimization problem for the projected subspaces, which was generally not considered in most of the existing soft subspace clustering algorithms. Maximizing the deviation of feature weights was proposed as the sub-space optimization goal, and a quantitative formula was presented. Based on the above, a new optimization objective function was designed which aimed at minimizing the within-cluster compactness while optimizing the soft subspace associated with each cluster. A new expression for feature-weight computation was mathematically derived, with which the new clustering algorithm was defined based on the framework of the classical k-means. The experimental results show that the proposed method significantly reduces the probability of trapping in local optimum prematurely and improves the stability of clustering results. And it has good performance and clustering efficiency, which is suitable for high-dimensional data cluster analysis.

As a new service for data storage and management, cloud storage has the virtue of portability and simplicity in use. However, it also prompts a significant problem of ensuring the integrity and recovery of data. A data recovery system for cloud storage based on fountain code was designed to resolve the problem. In this system, the user encoded his data by fountain code to make the tampered data recoverable, and tested the data's integrity with Hash functions so that the complexity in data verification and recovery was reduced. Through this system, the user can verify whether his data have been tampered or not by sending a challenge to the servers. Furthermore, once some data have been found tampered, the user can require and supervise the servers to locate and repair them timely. The experimental results show that the data integrity detection precision reaches 99% when the data's manipulation rate is 1%-5%.

Traditional n-gram feature extraction tends to produce a high-dimensional feature vector. High-dimensional data not only increases the difficulty of classification, but also increases the classification time. Aiming at this problem, this paper presented a feature extraction method based on Part-of-Speech (POS) tagging sequences. The principle of this method was to use POS sequences as text features to reduce feature dimension, according to the property that POS sequences can represent a kind of text.In the experiment,compared with the n-gram feature extraction, the feature extraction based on POS sequences at least improved the classification accuracy of 9% and reduced the dimension of 4816. The experimental results show that the method is suitable for emotion classification in micro blog.

There are limitations in establishing fuzzy logical relationship of the existing fuzzy time series forecasting methods, which makes it hard to adapt to the appearance of new relationship. In order to overcome the defects, an interval-similarity based fuzzy time series forecasting (ISFTS) algorithm was proposed. Firstly, based on fuzzy theory, an average-based method was used to redivide the intervals of the universe of discourse. Secondly, the fuzzy sets were defined and the historical data were fuzzified, then the third-order fuzzy logical relationships were established and a formula was used to measure the similarity between logical relationships. By computing the changing trend of future data, the fuzzy values were obtained. Finally, the fuzzy values were defuzzified and the forecasting values were obtained. The proposed algorithm makes up for the shortcomings in logical relationship of the existing forecasting algorithms because it forecasts the changing trend of future data. The experimental results show that the proposed algorithm ISFTS is superior to other forecasting algorithms on forecasting error, including Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). Therefore, the algorithm ISFTS is more adaptive in time series forecasting, especially in the case of large data.

Key point-based Symbolic Aggregate approximation (SAX) improving algorithm (KP_SAX) uses key points to measure point distance of time series based on SAX, which can measure the similarity of time series more effectively. However, it is too short of information about the patterns of time series to measure the similarity of time series reasonably. To overcome the defects, a composite metric method of time series similarity measurement based on SAX was proposed. The method synthesized both point distance measurement and pattern distance measurement. First, key points were used to further subdivide the Piecewise Aggregate Approximation (PAA) segments into several sub-segments, and then a triple including the information about the two kinds of distance measurement was used to represent each sub-segment. Finally a composite metric formula was used to measure the similarity between two time series. The calculation results can reflect the difference between two time series more effectively. The experimental results show that the proposed method is only 0.96% lower than KP_SAX algorithm in time efficiency. However, it is superior to the KP_SAX algorithm and the traditional SAX algorithm in differentiating between two time series.