Makine öğrenmesi yöntemleri ile el ve yüz analizi tahminlemesi = Hand and face analysis prediction with machine learning methods

Abstract

Individuals and institutions are increasingly concerned about the security of the technological infrastructure they use. Because security vulnerabilities in technological systems can cause serious harm and risks. The banking sector is in a particularly sensitive position in this respect. Banks must take strong security measures to protect their customers' financial assets and provide a safe environment. In our daily lives, many people apply for credit and make requests regarding the use of credit while performing their banking transactions. However, banks adopt a careful approach when evaluating loan applications and analyze the loan payment capacities of their customers. Part of these evaluations is examining the customer's personality characteristics. Banks attach great importance to personality analysis to predict the financial responsibilities and payment habits of their customers before providing loans. However, customers generally avoid disclosing their negative characteristics in loan applications. Therefore, it may be difficult to make an adequate assessment using traditional methods. In recent years, the use of artificial intelligence-supported methods to overcome such challenges has increased significantly. By analyzing large data sets, artificial intelligence can extract customer profiles in more detail and evaluate loan applications more effectively. For this reason, artificial intelligence-based facial analysis and personality analysis studies are attracting increasing attention in the banking sector. Detection and recognition of human faces is critical in security and surveillance fields. Facial recognition technology is developed through various methods used in this field. The main purpose of this technology is to detect and recognize human faces in an image and to identify them by comparing these faces with a predetermined database. In addition, various predictions can be made using facial recognition technology, such as analyzing the person's age, gender and even emotional state. Palm reading is a tradition that has a history in many cultures around the world and is used to predict future life. This art aims to inform and identify people by examining the lines, patterns and other features on the palm. This method, also known as palm reading or palm reading, is often associated with metaphysical or spiritual beliefs. On the other hand, palm print recognition is a biometric technique and is used to distinguish and verify the identity of individuals. The unique ridges, lines, and other features on the palm form during embryonic development and remain relatively unchanged throughout a person's life. The palm print created by combining these features provides an important differentiating feature between individuals. Therefore, palm print recognition is used as a reliable biometric recognition method in security systems, authentication processes and many other fields. This thesis focuses on topics such as face detection, face analysis and palm analysis, and examines and analyzes in depth the methods used in these fields. Studies in the literature are examined meticulously and which methods are used and what kind of results are obtained are determined in detail. The main purpose of the study is to research and develop effective methods that can be used in authentication systems based on face and palm analysis. In this direction, face and palm analysis in the developed system is used to accurately predict the person's features. In face analysis, information such as gender, age and emotion analysis of the person in the photo are estimated, while in palm analysis, predictions are made based on personality traits. The method of this study primarily utilizes the Python programming language and related libraries for face detection and analysis. Then, the data collection process required for personality analysis is initiated and this data is organized to include personality traits based on palm lines. Similar studies in the literature are examined in detail and analyzes are made on which methods are most effective. Next, a system is developed using the Python programming language. The user interface of the system is designed using a GUI library such as tkinter. This interface includes an area where the user can upload photos. If the uploaded photo contains a face image, the system analyzes the face and presents the output to the user. Machine learning libraries such as OpenCV and DeepFace are used for face analysis. OpenCv library is an open-source image processing library. The OpenCV library used is capable of detecting faces on images fed to our system. In this way, the images provided by the user as input are scanned and the facial images in them are detected. This detection process involves the process of identifying and determining facial features by analyzing the pixels of images. In this way, users can provide facial images to the system and perform various analyzes on these images. This feature is one of the key components of the project and is important to increase the functionality of the system. DeepFace library is an open-source facial recognition and analysis framework developed by Facebook. DeepFace library was used for face recognition and face analysis. The analyze function was used for face analysis. This function analyzes the features of the face contained in an image, such as gender, age and emotional state. The function takes the file path of an image as a parameter and predicts the features of faces in that image. As output, data containing values of specific features for each face is returned. By using machine learning techniques in palm analysis, palm analysis prediction was made on the obtained data set. These methods will allow the identification and analysis of personality traits associated with palm lines for the purpose of the research. Personality analysis is interpreted based on the life line, heart line and mind line data received from the user as input values. Necessary calculations were made with the Random Forest classification model, which is an effective machine learning algorithm for multiple classification problems. The RandomForestClassifier class is a classifier that implements the RandomForest algorithm. This algorithm makes a prediction by combining many decision trees and randomly subsampling. This model is used to predict whether input data corresponds to certain personality traits. Since the Random Forest algorithm is especially effective on complex data structures and high dimensional datasets, we proposed to use this algorithm. In this context, the accuracy rate was found to be %0.45. At the same time, calculations were made on the KNN machine learning algorithm with the obtained data set. As a result of this calculation, 38% MSE value was obtained. The low average accuracy value indicates that the model may not have learned the patterns in our data set well enough. Possible outputs of this study include valuable information presented to the user through accurately predicted gender, age, sentiment analysis and personality traits. This information plays an important role in authentication systems and personal analytics, helping users better understand their needs. This study adds a new dimension to research on face and palm analysis and provides a more effective method in the field of personality prediction by increasing the use of techniques in these fields.