IMPLEMENTING MACHINE LEARNING TO IMPROVE THE QUALITY AND SAFETY FRAMEWORK IN TANGKAP, A RIDE HAILING AND TECHNOLOGY COMPANY

• Main Author: Nugraha Saragih, Edo
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77447
• Institution: Institut Teknologi Bandung

The ride-hailing industry has existed in many places for decades. The rise of the internet and data revolution has only further added to fuel the rise of the industry. The industry has many players globally, starting from global brands spanning continents, regional power, and even local players. Each company has its own niche and value proposition. The uniqueness of each platform is a strong selling point they propose to get more value in the market. One important value proposition that many platforms aim to achieve is the safety and quality of each ride on their platform. The safety and Quality framework is important to ensure that each passenger of the ride-hailing experience a safe and enjoyable ride. The safe and enjoyable definition may vary by each person, but there are still bare minimums that each platform can achieve to ensure that people in their platform have a secure and enjoyable experience. Machine learning has been a surging technology in the past years. Machine learning is a technique where we teach a machine to learn from experiences. The teaching process includes using a set of training datasets with a recognizable pattern where the tool can repeat the process going forward. Machine learning tools have various models that apply to different uses. Some of the more common methods include BERT, LSTM, and Bag of Words. The usage of machine learning has been evident in different industries such as face recognition for platforms, sentiment analysis for media, or medical history classification. This research tries to implement machine learning tools in the ride-hailing company. The research aims to implement a machine learning tool to improve the safety and quality framework of Tangkap, a ride-hailing company. The machine learning tool that can be introduced for the Company is for the comment classification activity. Customer comments are important for ride-hailing companies to continue to improve their rides. Customer comments may be positive and full of compliments, but may also be full of complaints. These customer comments are then what ride-hailing companies used to treat their partners to improve each ride and booking on the platform. This research implements two types of machine learning models and tries to improve the processing time of comment classification in Tangkap. The two models this research implemented in Tangkap are LSTM (long short-term memory) and BERT (bi-directional encoder representation form transformer). Both LSTM and BERT have their advantages and disadvantages. LSTM is easier to train and has decent accuracy, but BERT should have more processing power but is much more expensive to train and take longer to complete processing. The research concluded that combining the two machine learning models with a manual eyeballing process at the end is the best option for Tangkap. The two models have different categorizations for some comments, and manual processing to check on the final categorization result from the two models is pertinent to create a stronger safety and quality framework for Tangkap. Both models will be hosted in Tangkap’s cloud storage and cloud computing tools. People in Tangkap can run the model to check on the daily comment categorization to ensure running as intended.