Crucial phases of testing, feedback integration, and continuous iteration #
External Feedback #
As part of improving the user experience of our web platform, an independent survey was conducted to gather feedback. The testing focused on evaluating the intuitiveness and usability of the interface, especially in the context of creating and editing student room allocations.
Overall impressions #
Most users expressed a high degree of satisfaction with their use of our platform. Several key aspects were noted:
- Interface clarity: Users found the interface clear and logical, making it easy to navigate.
- Intuitiveness of navigation: The ease of creating new allocations was emphasized, indicating efficient user interface design.
Editing allocations #
The page for editing current assignments was specifically tested:
- Allocation Process: Testers noted that the progress bar of assigning students to rooms was clear and did not cause difficulties.
- Page Adaptability: Despite changes in the state of the page during the editing process, users had no difficulty understanding the functionality.
Conclusions #
The overall external feedback indicates that the interface development is moving in the right direction. By paying attention to the details of user interaction, we can significantly improve the perception and usability of the platform.
Testing #
Our team is actively engaged in the development and improvement of machine learning and backend components. These efforts are aimed at improving the efficiency, reliability and scalability of our platform.
Machine Learning #
A novel metric has been devised for evaluation purposes. This metric’s valuation is computed through the average room ratio, the count of subscriptions, and the utmost feasible number of subscriptions within a designated area. An random distribution algorithm was selected as the foundational strategy. This algorithm acted as the foundation for subsequent experiments and analyses.
The examination encompassed three methodologies: spectral clusterization, spectral clusterization, and the approach utilized by the Louvain Institute. Each methodology was assessed relative to the benchmark set by the initially chosen random algorithm. Notably, all three algorithms exhibited a quadrupled - quintuple efficiency compared to the random algorithm. This enhancement was attributed to the algorithms’ inherent capacity to deliver more precise results.
Despite almost equal results for all three methods, the method of the Louvain Institute emerged as the superior alternative, albeit marginally, by a few percentage points. To augment the results further, the decision was made to integrate greedy logic into the evaluation procedure. Among the three algorithms scrutinized, this tactic permitted a minor but noteworthy boost in performance, elevating the metric by 5%. This improvement was considered substantial enough to justify further exploration.
Ultimately, the method of the Louvain Institute, enhanced with greedy heuristics, was selected for its potential to yield even superior results.
Backend #
At the moment of Sprint 4, the backend of the system is equipped with 450 unit tests. The emphasis on unit-testing is not accidental, because it is this type of tests that allows us to guarantee high quality and stability of the code at the earliest stages of development. Unit-tests provide checking the correctness of work of separate modules, which greatly simplifies error detection and elimination. Thanks to a thorough approach to testing, we can confidently move forward knowing that our system is reliable and ready for further development.
Iteration #
Continuous iteration plays a key role in the success of our project development. This process not only involves consistent product improvement through feedback and testing, but also involves close interaction and collaboration between different parts of the team. Our approach to iteration includes weekly team meetings to adjust plans and share experiences, which greatly enhances development efficiency.
Every week, our team meets to assess current progress and adjust the development plan. These meetings allow us to share experience and information between the frontend, backend and machine learning departments, which helps us to better understand the processes in each area.
The iterative process in our company is unique in that each development area is not only informed of the actions of other teams, but also actively participates in discussing and commenting on progress. This creates an environment for collaborative development, where each team member can contribute to the overall result.
By regularly evaluating and adjusting our plans and products based on a collective view and feedback, we not only adapt our products to changing requirements, but also continuously improve them in pursuit of our original goals.
Progress Report #
This week our team made significant progress on our key components - machine learning, backend and frontend.
You can learn about the progress of this sprint, as well as a detailed roadmap, on our page Notion Page
Challenges and Solutions #
Challenge A: The machine learning team encountered problems with the initial performance of the new recommendation algorithm, which did not meet the expected accuracy and efficiency metrics.
Solution: Using real-time performance data, we were able to improve our models more efficiently.
Challenge B: As we migrated the frontend design from Figma prototypes to a React implementation, maintaining visual and functional consistency across platforms and browsers was challenging.
Solution: The frontend development team has created a common component library that aligns with our system design principles. This library is integrated into our development process, ensuring that our user interfaces are consistent regardless of platform. We also conducted extensive cross-browser testing to ensure that our designs display correctly in all supported browsers.
Machine Learning #
The machine learning team has completed the development of the recommendation system algorithm, which has been successfully refined and validated. Future plans include work on distribution algorithms, which is the second type of algorithm we will be developing in the coming weeks.
Backend #
Development of the open source GraphQL API in Python continues, with plans to integrate the Telegram bot into the server in future sprints. This will be an important step in making our service more interactive and accessible.
Frontend #
In Frontend, the team started implementing the design from Figma on the Next.js framework. The main focus was on creating a responsive UI that adapts to different screen sizes and devices.
Next Steps #
Our next steps include:
- Machine Learning: Work on the distribution algorithm.
- Backend: Integration of the Telegram bot into the server.
- Frontend: Continue implementing the design from Figma.
As we continue to follow an iterative approach to development, we have maintained a high level of motivation and effective communication within the teams. This week was filled with important accomplishments that prepare us for the next phases of development. Our goal is to stay focused on the most significant aspects of the project and ensure its successful completion.