Mann-Whitney U test and Fisher's Exact test were both performed but not selected as the primary data analysis test. They are both good options for independent categorical variables testing. However, Mann-Whitney U test are good for samples with identical distribution. In this case, it is not clear and from the plot chart, the distribution are different (Boslaugh & Watters, 2008). Fisher's Exact test is appropriate for 2 by 2 contingency table. However, it normally substitutes Chi-square when samples are small and are sparsely distributed which is not this case (Boslaugh & Watters, 2008). 
Conclusions
Based on the results of Chi-square test, the null hypothesis can be rejected due to the very large Chi-square value. The alternative hypothesis is valid. Therefore, in August 2017,  it can be concluded that the proportion of short-term riders of all the riders during weekend is higher than the short-term riders during weekdays. However, in consideration of generalizing the assumption of more tourists tend to use Citibike during weekend in the summer, other months during other years need to be tests. In addition, weather condition needs to be considered while processing data. During flooding and thunderstorm season, the number of riders will fluctuate.  This mini-project assumes all short-term users are tourists. In reality, this assumption might not be easy to validate. 
References
1. Bianco, F.(2018). Effectiveness of NYC Post-Prison Employment Programs Solution. Retrieved from: https://github.com/fedhere/PUI2018_fb55/blob/master/HW5_fb55_session1/effectivenesofNYCPost-PrisonEmploymentPrograms_solution.ipynb
2. Boslaugh, S & Watters, P.A.(2008). Statistics in a Nutshell. O’Reilly Media, Inc.: Sebastopol, CA
3. Kvalnes, T. (2012). A Short Guide to the Choice of Statistical Tests.  Retrieved from: http://folk.ntnu.no/thomakva/teaching/Guide_to_statistical_tests.pdf
4. Stanford University. (2017). Comparing multiple proportions. Retrieved from: https://web.stanford.edu/class/psych10/schedule/P10_W7L1
Appendix
Fisher's Exact Test:
The data this Mini-project uses can be summarized into a 2 by 2 contingency table. As a result, Fisher's Exact Test was also performed to investigate the two samples as it is designed to comparing binary variables from contingency tables (Stanford University, 2017). The results shows that the p value is very small. Therefore, we can reject the null hypothesis.