The purpose of this project is to forecast average monthly temperature for Kraków, basing on provided observational dataset. Undoubtedly, machine learning is an ideal way to evaluate time series included this one.

The problem is to predict the average value of monthly temperature for Kraków. The dataset provides the monthly temperature from January 2007 to August 2018.

It is credited to Institute of Meteorology and Water Management, Poland.

The dataset has been downloaded as a CSV file with the filename "Krakow_2007_2018.csv".

• Jupyter Notebook - version 5.6.0

Jupyter Notebook documents are documents which contain both computer code and rich text elements. Notebook documents are both human-readable documents containing the analysis description and the results (figures, tables, etc..) as well as executable documents which can be run to perform data analysis.
(based on: www.jupyter-notebook-beginner-guide.readthedocs.io)

• Python - version 3.7

In order to get insight into project results open ARIMA.ipynb file in one's browser. However if one wants to execute the whole code step by step, the Jupyter Notebooks is required to be installed on one's computer. A significant annotation is that all those files (.ipynb, .py, .png, .csv) should be located in the same folder.

There are some advices, which can be helful to setup your local environement. The easiest way to get started with Jupyter Notebooks is by installing Anaconda.

1. To get Anaconda, simply download the latest version of Anaconda for Python 3 using this website . Install Anaconda by following the instructions on the download page and/or in the executable.
2. Open Git (if you do not have already installed it, click here) and clone the repo which contains ipynb notebook:

    $ git clone https://github.com/AleksandraPestka/Weather_forecasting/blob/master/ARIMA.ipynb

2. Go to the path where ipynb is cloned

3. Open Jupyter notebook to see the file.

    jupyter notebook ARIMA.ipynb

4. Run the code.

def evaluate_models(dataset, p_values, d_values, q_values):
    dataset_values=dataset.values
    dataset_values = dataset.astype('float32')
    best_score, best_cfg = float("inf"), None
    for p in p_values:
        for d in d_values:
            for q in q_values:
                order = (p,d,q)
                try:
                    mse = evaluate_arima.evaluate_arima_model(dataset_values, order)
                    if mse < best_score:
                        best_score, best_cfg = mse, order
                    print('ARIMA%s RMSE={:0.3f}'.format(order,mse))
                except:
                    continue
    print('Best ARIMA= {} RMSE= {:0.3f} '.format(best_cfg, best_score))

List of features ready:

• plotting a time series
• plotting seasonal line plots
• plotting density plot
• plotting box and whisker plots
• evaluating the best ARIMA model to the dataset
• reviewing residual errors
• making predictions for the new data

Project inspired by machine learning tutorials shared on the website: Machine Learning Mastery

Created by Aleksandra Pestka.