Methodology

The null hypothesis is the average of female Citi-bike users' trip duration is the same as male Citi-bike users. The two distribution is not a normal distribution so T-test or Z-test is not very compatible to do the hypothesis testing. Therefore, Mann-Whitney U test is chosen to do the test because Mann-Whitney U test can be performed on non-gaussian distributions. The Mann-Whitney U test is taking the rank of all data from 0 to the summation of the total number of sample sizes and comparing between two distribution of median of the ranks. 
Scipy.stats built-in function,  scipy.stats.mannwhitneyu(x,y), is used to calculate the p-value. The builtin function has a correction to a continuity of function and the result of p-value is 0.0. Therefore, the median of female and male trip duration doesn't have significant differences. 
 

Conclusion

Another Class mate comments on the hypothesis test.  Zhiao mentioned I should not take out outliners, however, the data set have huge standard deviation and cleaned out over 95 % of data is reasonable to me. Also, he suggests using Z-test however, the two distribution is not normal distribution so Z-test doesn't work. Federico Bianco commented on that my hypothesis equation was broken and fixed it. The p-value of Mann Whitney U test is 0.0 so the median of female and male trip duration doesn't have significant differences and I can't reject the null hypothesis.