Design and simulation of precision marketing recommendation system based on the NSSVD++ algorithm

This article is dedicated to discussing the design and improvement of precision marketing push algorithm under the deep learning target detection system. Scientific progress has made the network form a huge database, and each software platform can accurately locate the user’s market by analyzing the data, determine their interests, and carry out marketing pushes. Based on deep learning algorithms and models, simulation training can be performed on user rating data. It clusters users according to their preferences and performs data filtering to control the impact of data sparseness or differences between users. It uses the relevant similarity of users and uses neighborhood collaborative filtering to make accurate judgments and push. This article is based on the SVD++ (Singular Value Decomposition, SVD++) algorithm and optimized to achieve higher push accuracy. The previous explained the SDD (Single Shot MultiBox Detector, SDD) algorithm, Pearson correlation similarity, neighborhood-based collaborative filtering model, neural network model, and Rayleigh channel system to explain the application of the deep learning target detection system in precision marketing, then verify the feasibility of the improved SVD++ algorithm through experiments. In the experiment, through the comparative analysis of UserCF (Collaborative Filtering, UserCF), Slope one, SVD++, OrdRec, Pure, AllRank, JSVD++, MSSVD++ (Medical Society for the Study of Venereal Diseases, MSSVD++), NSSVD++ (Neighborhood Sampling Singular Value Decomposition) algorithms, the test focuses on JSVD++, MSSVD++, NSSVD++ algorithms. And it is concluded that the NSSVD++ algorithm, that is, the neighborhood sampling method, is the most effective and has the best marketing recommendation effect. Among them, the 1-Call algorithm is 41.62% higher than the SVD++ algorithm, and more than 16.88% higher than other benchmark algorithms, the COV (Covariance, COV) algorithm is improved by more than 12.97%, the CIL algorithm is improved by more than 16.12%, and the improvement of NSSVD++ is at least 71.37% based on the SVD++ algorithm. The experimental results show that the improved recommendation algorithm has better Top-N recommendation accuracy. Although there is indeed negative case information in the missing data, the direction of this experiment is the right direction. The results of the experiment have certain guiding significance for precision marketing push, and can achieve rapid development in this field.