Weather Forecast

From 10 years of weather data in 18 cities, the AI ​​is trained when: Enter time, location, and weather component, the output will show the prediction of climate composition at that time.

PowerPoint

Overview

• Data processing method: Interquartile Range (IQR)
• Approach to the problem (algorithm used): Model for data processing: Prophet: Open source library developed by Facebook, built on linear models combined with seasonal and event components, making forecasting easier should be easier and more accurate.
• Method of evaluating the problem: Mean Squared Error (MSE): average of the squares of errors between the predicted value and the actual value.

Installation

• Clone from github: Github
• Python, using pip to install libraries: pip install -r requirements.txt

Files

• Data: weather_prediction_dataset.csv
• Code: weather_forecast.py
• Python libraries: requirements.txt

Detailed Description

Development platform:

• Python: User-friendly, simple, readable syntax, with a strong community of machine learning tools.
• Google Colab: many libraries available, supporting error suggestions. Suitable for sharing with group members simultaneously.

Methods used:

• Data processing method: Interquartile Range (IQR): Interquartile range, used to identify and filter outliers values. A value is suspected to be an outlier if it is less than 1.5*IQR below the first quartile $(Q1 - 1.5\times IQR)$ or greater $(1.5 \times IQR)$ above the third quartile $(Q3 + 1.5 \times IQR)$.

→ This method assumes that data outside the upper and lower bounds are outliers

Approach: Prophet:

• Easy-to-use interface, easy to deploy.
• Weather data is data that changes periodically, → The tool is suitable for analyzing temporal data with clear seasonality.

Problem evaluation method: Prophet's MSE:

• Compare MSE of Arima and Prophet: ARIMA's MSE: Usually smaller when the data is linear, without many obvious nonlinear or seasonal factors. Prophet's MSE: Usually smaller in cases where the data has a nonlinear trend or clear seasonality.
• Prophet is often superior in more complex situations, while ARIMA may be better for simple time series.

Conclusion:

• Mean Squared Error on training set: 13.080

→ the model is already capable of capturing the characteristics of the data during the learning process. Mean Squared Error on test set: 13.057

→ Very close to the MSE on the training set,

→ Shows that the model is not overfitting and has good generalization ability when predicting on new data.