Skip to content

Vision and Scope

Jump to bottom
Khunakorn Pattayakorn edited this page Apr 15, 2025 · 7 revisions

Overview

ExtroPlanner, is a project designed to help event organizers plan outdoor events and activities and monitor the current environment.

Using historical weather data in Bangkok, ExtroPlanner aims to help organizers predict rainfall and harsh weather conditions to determine if an event should be conducted during the timeframe and what equipment they should prepare.

Data sources

Primary data source(s):

  • Temperature and humidity sensor (KY-015) for measuring real time temperature and humidity
  • CO sensor (MQ-9) for measuring CO levels

Secondary data source(s):

Scope

Main Features:

APIs to be provided to users

  • Calculated risk of heatstroke based on predicted weather conditions
  • Report current levels of CO in the area where the event is planned
  • Predicted rainfall and impact based on predicted weather conditions.
  • Descriptive statistics (Min-max and Avg) for weather conditions (temperature, CO, humidity/rainfall) in the area.
  • Predicted values for weather conditions in the area up to 1 week in advance.

Web Application:

  • Charts (line, pie, bar) to visualize historical and predicted weather data
  • Event advisor, to notify the impact of weather conditions such as:
    • Risk of heatstroke and current heat index.
    • Predicted rainfall during the event period.
    • Recommended items to prepare for the event.

Optional features for staged release

  • Real time forecasting for weather conditions 1 day in the future.
  • Support for weekly, monthly, and annual weather predictions.

Known constraints

  • Real time data is always changing, especially weather data which can shift very rapidly making weather predictions have an innate inaccuracy.
  • The initial data collected spans only a short period which can have noise, missing data, or local biases
    • Due to the small timeframe of the data it is impossible to account for annual seasonality.
    • Due to the daily seasonality of the data, predictions can only be based on an hourly basis.
    • Predictions for larger timeframes require a larger scale of data (daily averages over many months).
  • The pre-trained model is trained on a specific location, which may overfit the model on the location's data.
  • The MQ-9 sensor's sensitivity for CO ranges from 10-10000 ppm which makes it unfit for an early warning system.

Clone this wiki locally