Advertisement design in dynamic interactive scenarios using DeepFM and long short-term memory (LSTM)

Area of Science:

• Computer Science
• Artificial Intelligence
• Media Studies

Background:

• Traditional advertising evaluation methods for film and television face accuracy limitations in new media environments.
• Network-based media presents dynamic and interactive challenges for advertising data analysis.

Purpose of the Study:

• To propose a novel data-driven nonlinear dynamic neural network planning approach.
• To enhance the real-time evaluation precision and accuracy of film and television advertising in network media.

Main Methods:

• Developed a visual advertising design model for the dynamic interactive network media environment.
• Utilized DeepFM and long short-term memory (LSTM) modules within deep learning neural networks.
• Constructed an information statistics and data interest model using two public datasets and engineered features for visual advertising.

Main Results:

• The proposed visual neural network model demonstrated superior performance in capturing nonlinear dynamics and feature mining.
• Benchmarking against other models using F1 and root mean square error (RMSE) metrics confirmed the model's effectiveness.
• The model showed robustness and generalization capabilities in handling dynamic interactions between visual elements and text.

Conclusions:

• The data-driven neural network approach significantly improves advertising evaluation accuracy in interactive online media.
• The model's ability to process complex visual and textual data dynamics offers a robust solution for modern advertising challenges.