dliofindia/airtraffic_forecast

Creating Time Series Model for Air Flight Dataset

🔥 CREATING AIR FLIGHT PASSENGER FORECAST MODELS

✏️ PROJECT EXPLANATION

---

This project used Airport passenger of San Fransisco Airport from 2005 to 2016

Click here to get the Dataset

This Project was created to aim these following objectives :

• Create Time Series Model for Forecasting Purpose.
• Create Business Intelligence Dashboard .
• Create API for forecasting number of passenger of given period

✏️ PROJECT WORKFLOW

---

1. Dataset Preparation
2. EDA
3. Model Selection + Cross Validation
4. Model Result

🔧 Package / Tech Stack Used

---

• Time Series Model Development
  1. Statsmodels
  2. Pandas
  3. Numpy
  4. Neural Prophet
  5. sklearn
  6. xgboost

⌛ FUTURE PLANS

---

• Time Series Model Development
  1. Using LSTM, GRU, another complex model.
  2. Statistical Model like VAR
  3. Statspace Models : Kalman Filter, Hidden Markov Model

🔨 INSTALLATION

---

1. Clone the repo

#clone the repository first 
    git clone https://github.com/fakhrirobi/airtraffic_forecast.git

2. change directory

#change directory 
    pip install . 

3. Data Preparation

    cd airtraffic_forecast/data/prepare_data
    python generate_modified_data.py 

4. Begin Modelling

    cd airtraffic_forecast/modelling
    # --method : statistical, nn, ML 
    python start_modelling.py --method 

📗 Modelling Result

	RMSE Score	MAE Score	MAPE Score	model_name
1	2.9332e+10	120793	0.028163	RandomForestRegressor
2	3.94132e+10	160227	0.0383572	XGBRegressor
3	8.33301e+11	823894	0.195214	SVRegressor
4	1.98349e+11	364297	0.0889501	KNeighborsRegressor
5	1.51227e+11	323093	0.0806871	LinearRegression
6	3.45687e+11	484491	0.11898	PassiveAggressiveRegressor
7	4.29552e+09	52562.4	0.0160444	NeuralProphet_hidden_layer3_epoch_3_weekly_seasonality12
8	0.00879809	0.0750631	0.00494308	SARIMA_1,0,1_1,0,1,12_ts_log
9	0.000148183	0.00870418	0.0005716	SARIMA_1,0,1_1,0,1,12_ts_log_moving_avg
10	3.51627e+10	127422	0.0310854	SARIMA_1,0,1_1,0,1,12_ts_moving_avg
11	0.000148428	0.00895553	0.000589953	SARIMA_1,0,1_1,0,1,12_ts_log_ewma
12	0.00758013	0.0672904	0.977108	SARIMA_1,0,1_1,0,1,12_ts_log_ewma_diff
13	7724.52	69.9086	0.0352462	SARIMA_1,0,1_1,0,1,12_sqrt_ts
14	308.334	10.7121	0.00528588	SARIMA_1,0,1_1,0,1,12_moving_avg_sqrt

..

📕 ODDS

---

Dashboard Development :

1. Statsmodels SARIMAX syntax is not similar to sklearn .fit(data) but instead the data is defined when defining the model

..

📧 Connect With Me

---

1. Linkedin
2. Medium
3. Kaggle