This paper examines racial bias and discriminations in machine learning algorithms using America’s longitudinal high school students dataset. This study reveals machine learning algorithms may present a seemingly fair accuracy for both White and Asian student group and Black and Hispanic student group, but underneath the surface, the machine learning algorithms consistently produce a higher false positive rate for the White/Asian student groups while it consistently underestimates Black/Hispanic student group’s 12th grade math performance. This paper provides a comprehensive analysis and comparison of seven commonly used machine learning algorithms’ performances in terms of biased results towards the White and Asian student groups versus Black and Hispanic student groups.