Time Series Analysis Projects

• Project 1: Comparing time series forecasting of COVID-19 deaths

  • Task 1: Preprocess the data using pandas to be ready for machine learning, and visualize the data using matplotlib
  • Task 2: Create a SARIMAX model, optimize the model hyperparameters, use the model for forecasting future COVID-19 deaths and visualize the results
  • Task 3: Create a prophet model and use the model for forecasting future COVID-19 deaths and visualize the results
  • Task 4: Create a function that extracts features for training the XGBOOST and a feedforward neural network models
  • Task 5: Split time series feature dataset into training and test datasets and perform data normalization
  • Task 6: Train an XGBOOST model and a feedforward neural network model, and finally compare the predictions of all the models covered in the project

• Project 2: Comparing time series forecasting of pedestrian/cyclists counts

  • Setting up the data set and performing EDA
  • Identifying Stationarity
  • Transform Data to be Stationary
  • Triple Exponential Smoothing
  • Forecasting Using SARIMA model
  • Forcasting using Facebook's Prophet model
  • Forcasting using LSTM/GRU model
  • Forcasting using XGBoost

Time Series Analysis Courses

• Time series data analysis with Pandas library

  • Understand time series applications for NumPy and Pandas
  • Summarize a dataframe with a datetime index
    • setindext('date') for accessing Datetime Components
    • reindex and daterange for Handling duplicate or missing indices
    • resample for upsampling (e.g. moving from Monthly to Annual) and downsampling (e.g. moving from Annual to Monthly)
  • Generate simple time series plots
    • pct_change() to get Variable Percent Change
    • rolling(window_size) to get rolling mean/STD
    • Time series plots from statsmodels.graphics.tsaplots:
      • plot_acf: Plot of the Autocorrelation Function
      • plot_pacf: Plot of the Partial Autocorrelation Function
      • month_plot: Seasonal Plot for Monthly Data
      • quarter_plot: Seasonal Plot for Quarterly Data

• Decomposition of time series data into three components (Trend, Seasonality, Residual)

  • seasonal_decompose()

• Assessing stationarity of time series data sets and transformations methods

  • What it means for time series to be stationary.
    • constant mean, constant variance, constant autocorrelation structure, no periodic component
  • Common ways to identify stationarity.
    • run_sequence_plot()
    • calculate statistics for each splitted chunk of data (np.split())
    • Histogram
    • Augmented Dickey-Fuller Test adfuller
  • Useful nonstationary-to-stationary transformations.
    • Decomposition models for removing trend/seasonality
    • Log-transformation
    • Removing autocorrelation with differencing by approporiate lag

• Simple and exponentially weighted moving average smoothing of time series

• Developing Autoregressive-Moving Average (ARMA) models

  • A practical understanding of Autoregressive (AR) models.
  • A practical understanding of Moving Average (MA) models.
  • A practical understanding of the Autocorrelation Function (ACF).
  • A practical understanding of the Partial Autocorrelation Function (PACF).
  • Insight into choosing the order q of MA models.
  • Insight into choosing the order p of AR models.

• Developing Autoregressive Intergrated Moving Average (ARIMA) using SARIMAX and FaceBook Prophet

  • A practical understanding of ARIMA models
  • Insight into checking fit of model
  • Create forecasts with ARIMA models
  • A practical understanding of fbprophet
  • How to check fit of fbprophet model
  • Means of adjusting and improving 'fbprophet' model parameters

• Train/Test RNN, LSTM and GRU models for time series forecasting

  • Build and train RNN, LSTM, GRU for time series forecasting, using keras
  • Fine-tune RNN/LSTM parameters