In recent years, the employment problem is a social problem that needs to be solved urgently. With the rapid development of network technology, the trend of human resources networking is gradually increasing, and human resources recommendation faces the problem of information overload. Traditional recommendation algorithms cannot adapt to the phenomenon of information expansion in the era of big data. A large amount of job information makes job seekers easily fall into information review fatigue, so that they cannot clarify their job search needs, which leads to more energy to find the one that really suits them job information. This paper aims to study the application of human resource recommendation algorithms relying on decision trees. This paper takes human resources recommendation as the application scenario, and propose an improved human resources recommendation algorithm. The main work includes the following points: (1) Research and implement streaming distributed Data collection technology is used to collect job seeker information, job position information and user behavior information. Combined with the characteristics of human resources, the collected data is preprocessed such as data cleaning, data extraction and data conversion. (2) A human resources recommendation algorithm combining is proposed. First, the feature conversion ability of gradient boosting tree is used to complete the screening and encoding of original features, and then the converted features are input into the design of this paper. The hybrid convolutional neural network uses convolution operations for high-level feature learning, and realizes personalized human resources recommendation.