Algorithm design and analysis ability reflects the basic abilities of computer science students, and autonomous learning ability is related to their development potential. Research and develop Project-Based Blended Learning (PJBBL） Mode to change traditional learning methods and improve the algorithm design and analysis ability and autonomous learning ability of computer science students.The PBBL model has been revised multiple times using methods such as literature research, teacher needs analysis, and focus group discussions.This model includes six stages: build project , scenario construction, divisions of student roles, project plan, project implementation, project Evaluation and improvement. all of which integrate blended learning both online and offline. Experiments were conducted to verify the effectiveness of the PBBL model.The experimental group comprised 36 participants, while the control group consisted of 33 participants. The algorithm design and analysis course learning in the experimental group utilized the PJBBL model, whereas the control group followed the traditional learning model. Following the completion of the experiment, achievement were administered to assess the algorithm design and analysis abilities of both the experimental and control groups. Subsequently, an evaluation of autonomous learning ability was conducted, and data analysis was performed using MANOVA. Levene's test revealed that the Sig. of the achievement in both the experimental and control groups was 0.104, while the Sig. of autonomous ability was 0.066, with both values exceeding 0.05, indicating similarity in variance between the groups. The achievement MANOVA was a significance value of 0.005, and the autonomy ability score was 0.000, both of which were less than 0.05. These results indicated a significant difference between the two dependent variables, achievement and autonomous ability, when comparing the experimental and control groups.