data_analysis_3c.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Mon Sep 11 01:25:40 2023

@author: st_ko
"""

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from functools import reduce
import os
import keras as ks
from utils import *
import tensorflow as tf





# function to plot resampled momthly (mean of all month) observations for all years for all currencies
# for each currency draw different plot
def plot_all_currencies_monthly(data,size,num_currencies,ann=False,save=False):
    # annotate a subpart of the series
    for col in data.columns[0:num_currencies]:
        fig=plt.figure()
        for i in range(size):
            if(ann):
                plt.annotate(data[col][i], (data[col].index[i], data[col].values[i]),
                         textcoords="offset points", xytext=(0, 10), ha='center',fontsize=8,rotation = 45)
        plt.plot(data[col][:size],'o-',label=col)
        plt.legend(loc='lower right',bbox_to_anchor=(1.05,0.2))
        plt.title("Monthly sampled All years " + col )
        if(save):
            # 1000 means all years but monthly
            # just a random chosen number
            save_image(fig, col,None)
            plt.close()



# plot the currencies within a given range of observations
# added option to select currencies subrange as well
# this plots chosen subset of series and observations all together in one plot
def plot_Series(data,size,num_currencies,ann=False):
    # annotate a subpart of the series
    plt.figure()
    for col in data.columns[0:num_currencies]:
        for i in range(size):
            if(ann):
                plt.annotate(data[col][i], (data[col].index[i], data[col].values[i]),
                         textcoords="offset points", xytext=(0, 10), ha='center',fontsize=8,rotation = 45)
        plt.plot(data[col][:size],'o-',label=col)
        plt.legend(loc='lower right',bbox_to_anchor=(1.05,0.2))



# ! More cuztomized
# create a nested dictionary of year -> dictionary of months --> observations of this month in this year

def create_nested_dict(data):
        total_time= [x for x in data.index]
        year_dict = {x: {x2 : [] for x2 in iter(range(1,13)) } for x in range(data.index.min().year,data.index.max().year+ 1)}
        m_step = data.index.min().month
        y_step = data.index.min().year
        y=True
        m = True

        # create nested dictionary
        # i will turn this later into a function
        for i in range(len(total_time)-1) :
            if(total_time[i].year == total_time[i+1].year):
                y = True
                if(total_time[i].month == total_time[i+1].month):
                    year_dict[y_step][m_step].append(total_time[i])
                    m = True
                else :
                    if(m==True):
                        year_dict[y_step][m_step].append(total_time[i])
                        m=False
                    m_step +=1
            else :
                # reset months to set the last observation
                m_step = 12
                # add the remaining same year observation
                year_dict[y_step][m_step].append(total_time[i])
                # reset month index
                m_step=data.index.min().month
                y_step+=1
                y=False
        return year_dict



# for a certain year , for a certain currency plot the monthy series of this currency (one subplot per month of the year)
def plot_one_year_one_currency(year_dict,data,year,currency,save=False):
    # total plots
    fig, axes = plt.subplots(nrows=3, ncols=4, figsize=(15, 10))  # 3x4 grid for subplots
    #select currency from dataset


    # for each month
    for month in range(1, 13):
        row = (month - 1) // 4  # Determine row index
        col = (month - 1) % 4   # Determine column index

        #pick year and currency
        dollar_month = data.loc[year_dict[year][month],currency]
        ax = axes[row,col]
        # rotate txt
        ax.xaxis.set_tick_params(rotation=60)
        ax.plot(dollar_month,'o-',label= currency + ' : ' + str(month))
        ax.set_title(f"{year}-{month}-{currency}")

    #general title for whole plot
    plt.suptitle(f"Time Series Observations for Year {year}", fontsize=16)
    plt.tight_layout()
    #plt.show()
    # close plot
    #plt.close()



    # save image (optional)
    if(save):
        save_image(fig,currency,year)


# same as above but do it for all currencies (17 currencies -> 17 plots )
def plot_one_year_all_currencies(year_dict,data,year,save=False):
    for cu in data.columns:
        plot_one_year_one_currency(year_dict,data,year,cu,save)



# plot all years one currency
def plot_all_years_one_currency(year_dict,data,currency,save=False):
    # total plots
    fig, axes = plt.subplots(nrows=5, ncols=5, figsize=(15, 10))  # 3x4 grid for subplots
    #select currency from dataset


    # for each month
    inyear = data.index.min().year
    for year in range(1, 26):
        row = (year - 1) // 5  # Determine row index
        col = (year - 1) % 5   # Determine column index

        # take the yearly indices
        indices = [k for k2 in [x for x in y_dict[inyear].values()] for k in k2]


        dollar_month = data.loc[ indices ,currency]
        ax = axes[row,col]
        # rotate txt
        ax.xaxis.set_tick_params(rotation=60)
        ax.plot(dollar_month,'o-',label= currency + ' : ' + str(inyear))
        ax.set_title(f"{inyear}-{currency}")
        inyear += 1

    #general title for whole plot
    plt.suptitle(f"Time Series Observations for All years - {currency}", fontsize=16)
    plt.tight_layout()
    #plt.show()
    #plt.close()

    if(save):
        # i use 25 to denote that we plot all the years
        save_image(fig,currency,25)





# function to plot all years for all currencies
def plot_all_years_all_currencies(year_dict,data,save=False):
    for cur in data.columns:
            plot_all_years_one_currency(year_dict, data, cur,save)




# function to save images
def save_image(fig,currency,year):
    if not (os.path.exists('plots')):
            os.makedirs('plots')
    plt.savefig(os.path.join('plots', currency + ' ' +str(year) +  '-forecast.png'))
    plt.close()






# only called as main for testing
if __name__=="__main__" :

        pass