In the era of big data, artificial intelligence (AI) has been widely applied in higher education, providing technical supports to practical teaching in colleg-es and universities. This paper mainly creates an AI-based practical teaching mode for cultural industry management major of Chongqing Three Gorges University. Firstly, an intelligent management cloud platform was established for practical teaching, drawing the merits from Massive Open Online Course (MOOC) and Self-Paced Open Course (SPOC). Meanwhile, the AI technique was adopted to realize personalized learning and provide intelligent push services. In this way, the online MOOC+SPOC platform seamlessly integrates the teaching content into specific teaching scenarios, and the offline cloud platform manages the teaching process in an intelligent manner. Under the proposed teaching mode, the teaching content could match the job compe-tence standards and the actual abilities of college students. The research re-sults provide a new mode of practical teaching that covers all dimensions and promotes personalized and collaborative learning.

Rain classroom is a big data tool that effectively connects the teacher with students throughout the teaching process. This paper mainly applies rain classroom in blended teaching of college students, and evaluates the application effect. Firstly, the authors set up a model of rain classroom, covering all three phases of the teaching process: before-class (B), in-class (I) and after-class (A). Next, the BIA model was applied to the course Film and Television Appreciation, and the key issues in each phase were explained. To evaluate the effect of the BIA model, two questionnaire surveys were carried out among engineering students in Chongqing Three Gorges University. The results show that rain classroom can greatly improve the learning effect of the target course in various aspects: the teacher could arouse the students’ learning interest by sending red packets, make students more attentive through limited-time quiz, and reduce the absence through random roll call; the students were actively involved in group activities and confident in presenting their findings; however, many students most students switched to other apps in the class. The research results provide new insights to the application of big data technology in college education.

Some learners couldn’t get ideal learning results from distance education, to find out the root of this problem, their learning behavior and initiative should be analyzed in real time. However, in existing research results, the collected distance learning behavior data of learners are not pertinent enough, the extraction methods of big data are not proper enough, and the analysis models of learning behavior are not scientific enough. For these reasons, this paper analyzed the autonomous learning behavior of learners during distance reading based on big data, and gave the analysis contents and methods. In the text, a sequence diagram transformation method that is self-adaptive to long and short sequences had been introduced to process the learning behavior data of learners, who were then classified according to the features of different autonomous learning behavior sequences. Then, an attribute reduction algorithm based on improved Bayesian fuzzy rough set was adopted for attribute reduction, and the behavior indexes that are closely related to the autonomous learning effect of learners were selected for correlation analysis. After that, this paper proposed a method for detecting non-autonomous learning behavior based on multiple time scales, combining short-term autonomous learning pattern with long-term autonomous learning behavior and resource access behavior, this paper also analyzed the typical autonomous learning pattern of learners through the learning of hidden layer features. At last, experimental results proved the effectiveness of the proposed analysis method.

Background: On the retrieval of spatiotemporal information of chorography (STIC), one of the most important topics is how to quickly pinpoint the desired STIC text out of the massive chorography databases. Domestically, there are not diverse means to retrieve the spatiotemporal information from chorography database. Emerging techniques like data mining, artificial intelligence (AI), and natural language processing (NLP) should be introduced into the informatization of chorography. Objective: This study intends to devise an information retrieval method for STIC based on deep learning, and fully demonstrates its feasibility. Methods: Firstly, the authors explained the flow for retrieving and analyzing the data features of STIC texts, and established a deep hash model for STIC texts. Next, the data matching flow was defined for STIC texts, the learned hash code was adopted as the memory address of STIC texts, and the hash Hamming distance of the text information was computed through linear search, thereby completing the task of STIC retrieval. Results: Our STIC text feature extraction model learned better STIC text features than the contrastive method. It learned many hash features, and differentiated between different information well, when there were many hash bits. Conclusion: In addition, our hash algorithm achieved the best retrieval accuracy among various methods. Finally, the hash features acquired by our algorithm can accelerate the retrieval speed of STIC texts. These experimental results demonstrate the effectiveness of the proposed model and algorithm.