This research article proposes a method for predicting autism traits using attention-focused eye gaze analysis and transfer learning. The study focuses on the importance of eye positioning as a key feature for identifying autism. The research is conducted in two phases. In the first phase, various transfer learning algorithms are implemented and evaluated on open-source image datasets, Kaggle and Zenodo, to predict autism traits. Fivefold cross-validation is used to reinforce the results. The ConvNextBase model achieved the highest diagnostic accuracy, with 80.4% on Kaggle and 80.71% on Zenodo. The model uses a dlib library with HOG and Linear SVM-based face detectors to identify facial parameters, specifically the Eye Aspect Ratio (EAR), to measure attentiveness. If the EAR value is less than 0.20 for more than 100 consecutive frames, the model concludes the participant is un-attentive. The model generates a graph to visualize the participant's attentiveness over time. The second phase of the study focuses on using attentiveness to identify autism traits. The model uses the EAR value to determine if a participant is attentive or disengaged. The study highlights the importance of attention as a parameter for diagnosing visual behavior and predicting autism traits. The research also discusses the challenges of non-clinical diagnosis, including time-consuming and costly processes. The study emphasizes the need for efficient and accurate methods to diagnose autism traits. The proposed model uses deep learning algorithms, including transfer learning, to analyze eye gaze data and predict autism traits. The study also discusses the limitations of existing methods and the potential of transfer learning in improving the accuracy and efficiency of autism diagnosis. The research contributes to the field of autism diagnosis by providing a novel approach using attention-focused eye gaze analysis and transfer learning.This research article proposes a method for predicting autism traits using attention-focused eye gaze analysis and transfer learning. The study focuses on the importance of eye positioning as a key feature for identifying autism. The research is conducted in two phases. In the first phase, various transfer learning algorithms are implemented and evaluated on open-source image datasets, Kaggle and Zenodo, to predict autism traits. Fivefold cross-validation is used to reinforce the results. The ConvNextBase model achieved the highest diagnostic accuracy, with 80.4% on Kaggle and 80.71% on Zenodo. The model uses a dlib library with HOG and Linear SVM-based face detectors to identify facial parameters, specifically the Eye Aspect Ratio (EAR), to measure attentiveness. If the EAR value is less than 0.20 for more than 100 consecutive frames, the model concludes the participant is un-attentive. The model generates a graph to visualize the participant's attentiveness over time. The second phase of the study focuses on using attentiveness to identify autism traits. The model uses the EAR value to determine if a participant is attentive or disengaged. The study highlights the importance of attention as a parameter for diagnosing visual behavior and predicting autism traits. The research also discusses the challenges of non-clinical diagnosis, including time-consuming and costly processes. The study emphasizes the need for efficient and accurate methods to diagnose autism traits. The proposed model uses deep learning algorithms, including transfer learning, to analyze eye gaze data and predict autism traits. The study also discusses the limitations of existing methods and the potential of transfer learning in improving the accuracy and efficiency of autism diagnosis. The research contributes to the field of autism diagnosis by providing a novel approach using attention-focused eye gaze analysis and transfer learning.