General Overview of the Project

This project attempts to develop a tool which classifies the complaints written to Istanbul Metropolitan Municipality with respect to the municipal enterprise they are concerned with. We aim for increased efficiency in complaint handling for the municipality and quicker municipal response for the citizen complaints.

Main Findings

We set f1-score as the main indicator of the performance of our algorithms. For all 4 algorithms, we have obtained f1-scores above 90%, indicating the reliability of predictions. Accuracies obtained, further corroborate our confidence in the algorithms.


Figure 1: f1-scores of 4 algorithms used


Figure 2: Accuracies of 4 algorithms used


Conclusion and Policy Implications

Given the f1-scores and accuracies above, the project has the potential to create tangible benefit to the municipality. While we created labels based on the complaints available on the Internet, according to more specific needs and priorities of the municipality, this project can be replicated with the different labels.

In the more distant future, the municipality may adopt a speech recognition mechanism for the hotline ALO 153. After citizens summarize their complaints with a few key words over the phone, the algorithms may detect the enterprise responsible for the complaint and redirect the citizen to that enterprise.

References

[List your references here using APA format (or at least standardize all of the references using the same format).]