@zico
on
Habit completion predictor
4h 7m 3s logged
@zico
on
Habit completion predictor
4h 7m 3s logged
@zico
on
Habit completion predictor
1h 26m 2s logged
Sorry i haven’t been actively posting a devlog.
I’m currently working on the UI of the project as it got sent back after been reviewed. While testing the application, i found out some core issues like login in taking more time to verify and SQLAlchemy reusing closed database session pool.
I’ve debugged and optimize the logging time as best as i can, although the main blocker is my server (Render) which puts the server to sleep after a while and wakes back up when a request is been sent to it which also influences the server response time but not necessarily the main cause as bcrypt password verification also takes more computational resources and database connection.
For the 2nd bug fix i made sure that before any query gets through, it runs a pool_pre_ping to verify if it’s still active, else it would replace the stale database connection with a new one
Ship
Pending review
@zico
on
Habit completion predictor
I just wrapped up the development of the prediction model although i’m yet to implement an auto retraining of model based on users data because i have little data to actually analyse which model would be the best which still leaves the model at the default one trained with synthetic data. Your data are been kept tho and would be trained on when the feature is active and your logs is up to the min requirement
@zico
on
Habit completion predictor
3h 53m 54s logged
I just fixed a server failure for the backend due to duplicate db table. It’s up and running to use now. I’ll be improving feature engineering for the underlyin ML Model after i’ve gathered enough real data. Feel free to check out the project and suggest any changes 😊
@zico
on
Habit completion predictor
36h 21m 46s logged
Successfully trained the model with an accuracy of 99% (PSS: That’s because i used synthetic dataset and it wasn’t realistic enough).
What i’ve done so far:
- Feature Engineering: Mapped out features that could be useful in predicting an habit completion and filtered out noise features
- Created a synthetic dataset for the ML training using Logistic Regression and Decision Tree (Both seemed to work well tho).
- I tested the ML model with the training dataset and did a mini visual evaluation on the model accuracy and features dependency.
Ship
I created a chess-engine with a chess bot that plays smarter move by looking into the game by 3 steps or even more.
Managing states was challenging to me at first cause the state was mutated which caused lot of error and missing piece so i cloned the board which makes it immutable.
I’m proud because this project taught me a lot of things from state management to writing my first algorithm.
You can check out my project at https://zico-chess-engine.vercel.app. My repo at https://github.com/young1dev/chess-engine
7h 50m 17s logged
Added the min-max and alpha-beta pruning algorithm to enable the bot make smarter moves
2h 53m 30s logged
I’m working on the v2 of my chessbot using the minmax algorithm and alpha beta pruning
