Multi-Arm-Bandits-for-Advertisement-click-through-rate-optimization

Advertisement optimization refers to the process to categorizing advertisements specific to a certain group of the population. A few well-known techniques for advertisement optimization include A/B testing and greedy algorithms. This project deals with analyzing the performances of techniques such as epsilon greedy algorithm, Upper Confidence Bound (UCB) algorithm and Thompson Sampling. Experiments conclude that Thompson sampling outperforms the rest of the above-mentioned algorithms. However, when taking hyper-parameter tuning into consideration, epsilon greedy algorithm results in the lowest regret.