Research And Application Of Machine Learning On Danxia Information System

In the big data era,an information system that is able to flexibly scale out,store mass data,quickly response to concurrent requests,and provide intelligential services based on the data stored is particularly important.Despite the mature mining technologies on structured data,the utilization of images and other unstructured data is still inadequate which results in the waste of data sources.Under this circumstance,this paper analyzes the DanXia landform data and system requirements,builds a scalable information system to which machine learning technologies are applied providing users with practical functions like DanXia landform photo recognition and DanXia landform recommendation.This project uses neural network technology to support that users can retrieve DanXia landform data with image input.Since fully connected neural networks get too many free parameters and low training efficiency when coping with high-resolution images classification problem,this paper design a convolutional neural network architecture that is able to classify high-resolution DanXia landform image efficiently.The system regularly crawls the latest DanXia landform images from the database and generates training data set of the convolutional neural network after certain preprocessing.Once the training is complete using gradient descent algorithm,users can upload a picture and get the prediction from the system about which DanXia landform the picture belongs to.The system recommends appropriate DanXia scenic to users by taking advantage of collaborative filtering.However,when dealing with DanXia scenic recommendations the classic collaborative filtering algorithm has the drawback that it ignores the external influence of rating which leads to low recommendation accuracy.In view of this challenge,an optimized collaborative filtering algorithm is put forward in this paper.Concretely,the algorithm takes date time into account when decomposing factors and predicting users' ratings.After quantifying the impact of external factors and extending free parameters of users,a new cost function of collaborative filtering is proposed.Once the training is complete,users can get better DanXia scenic recommendations according to the date time they enquire.After building a scalable information systems and applying machine learning technologies to the DanXia landform data,a series of experiments are conducted demonstrating the success of machine learning technology application and unstructured data utilization.This paper also provides guidance to researchers who build scalable information system and apply machine learning technology on similar characteristic data.