As a useful tool, probabilistic hesitant fuzzy set is an enhanced version for hesitant fuzzy set. It could be used to model the uncertainty very effectively. However, in probabilistic hesitant fuzzy risky multiple attribute decision making problems, the occurrence probabilities of elements in a probabilistic hesitant fuzzy element and the probability of risk status are often difficult to obtain by subjective evaluation of a decision maker. This paper aims to propose two nonlinear programming models for calculating the probabilities of elements in a probabilistic hesitant fuzzy element and the probability of risk status respectively. First, a nonlinear programming model using maximum entropy principle is established for determining the probabilities of elements in a probabilistic hesitant fuzzy element. Second, by introducing the water-filling theory, we put forward its extension and design a novel mathematical programming model to determine the probability of risk status. Moreover, we have proved that both the two mathematical programming models are convex programming models and their global optimal solutions can be found. Thirdly, the collective overall expected values of alternatives are calculated and the ranking order can be derived. Then, the selection of investment project is investigated, and comparison analysis shows the superiority of the presented approach.

Short text is becoming ubiquitous in many modern information systems. Due to the shortness and sparseness of short texts, there are less informative word co-occurrences among them, which naturally pose great difficulty for classification tasks on such data. To overcome this difficulty, this paper proposes a new way for effectively classifying the short texts. Our method is based on a key observation that there usually exists ordered subsets in short texts, which is termed ``information path'' in this work, and classification on each subset based on the classification results of some pervious subsets can yield higher overall accuracy than classifying the entire data set directly. We propose a method to detect the information path and employ it in short text classification. Different from the state-of-art methods, our method does not require any external knowledge or corpus that usually need careful fine-tuning, which makes our method easier and more robust on different data sets. Experiments on two real world data sets show the effectiveness of the proposed method and its superiority over the existing methods.