Metals_Prices_Prediction

Metal Price Prediction Business case study:

• This was a famous bet between Julian Simon(Business professor) and Paul Ehrlich(Economist) from 1980 to 1990.
• Julian simon challenged paul Ehrlich to pick any five metals in between 1980 and 1990. simon told that metals price goes down even though
• inflation is there because technology will catch up to match the demand of metals thus not letting the metal price to go high
• on the other hand, paul told metal prices will raise sky high because of inflation when calculated in real terms.
• The bet was formalized in 1980, with 1990 as the payoff date.

Packages to be installed:

        library(pacman)
        pacman::p_load(forecast)
        p_load(tidyverse, lubridate, rio, pdfetch, tidyverse, readxl,dygraphs)
        library(dygraphs)
        
                # Secondary packages        
                pacman::p_load(tsibble, fable)
                pacman::p_load(tsibbledata)
                pacman::p_load(feasts, fpp3)
                library(ggthemes)
                library(tsbox) #convert xts to tsibble
                library(TSstudio)
                library(xts)
                library(zoo)
                library(ggthemes)

Import and preview the data

tungsten <- import("https://pubs.usgs.gov/sir/2012/5188/tables/tungsten.xlsx", skip = 2)

## New names:
## * `` -> ...2

head(tungsten); tail(tungsten)

##   Year ...2 Price
## 1 1959   NA    13
## 2 1960   NA    19
## 3 1961   NA    17
## 4 1962   NA    12
## 5 1963   NA     9
## 6 1964   NA    15

##                                                                                                                                                                                                                   Year
## 54                                                                                                                      Note: Annual average prices were derived from price changes reported in the following sources:
## 55                                                                                                       1959–66, tungsten ore (wolframite) in New York, “ordinary quality,” excluding duty, in American Metal Market.
## 56                           1967–73, tungsten ore, domestic quote reflecting the U.S. Government's General Services Administration price, in American Metal Market's Metal Statistics 1972 and Metal Statistics 1974.
## 57 1974–76, tungsten ore, minimum 65% tungsten trioxide, European market, excluding duty, in U.S. Bureau of Mines Minerals Yearbook, converted from pounds sterling per metric ton unit as reported in Metal Bulletin.
## 58                                                                                                      1977–88, tungsten ore, minimum 65% tungsten trioxide, U.S. spot price, c.i.f., excluding duty, in Metals Week.
## 59                                                                                                                                             1989–2010, ammonium paratungstate, U.S. free market, in Metal Bulletin.
##    ...2 Price
## 54   NA    NA
## 55   NA    NA
## 56   NA    NA
## 57   NA    NA
## 58   NA    NA
## 59   NA    NA

tungsten[1:2] = NULL  #This step will remove the first and second columns because we need only one vector
str(tungsten)

## 'data.frame':    59 obs. of  1 variable:
##  $ Price: num  13 19 17 12 9 15 23 38 43 43 ...

Data Preparation

• Remove the null values

tungsten = na.omit(tungsten)

Convert the vector to the time-series variable

tungsten <- ts(tungsten, start=c(1959), end=c(2010), frequency=1)

Cpi Index:

• This helps to convert nominal prices of metals to real prices.
• There are lot of indexes available online . Based on the users and company’s needs
• The perfect index has to be picked.

cpi = pdfetch::pdfetch_FRED("CPIAUCSL")  #download cpi all
              head(cpi)

##            CPIAUCSL
## 1947-01-31    21.48
## 1947-02-28    21.62
## 1947-03-31    22.00
## 1947-04-30    22.00
## 1947-05-31    21.95
## 1947-06-30    22.08

                tail(cpi)

##            CPIAUCSL
## 2021-04-30  266.832
## 2021-05-31  268.551
## 2021-06-30  270.981
## 2021-07-31  272.265
## 2021-08-31  273.012
## 2021-09-30  274.138

                  str(cpi)

## An 'xts' object on 1947-01-31/2021-09-30 containing:
##   Data: num [1:897, 1] 21.5 21.6 22 22 21.9 ...
##  - attr(*, "dimnames")=List of 2
##   ..$ : NULL
##   ..$ : chr "CPIAUCSL"
##   Indexed by objects of class: [Date] TZ: UTC
##   xts Attributes:  
##  NULL

            cpi.ts = ts(cpi, start = c(1947,1), end = c(2021,7), frequency = 12) 

Match the time-interval of tungsten and index.

• Both the nominal prices of metal and the index prices must start and end on the same time interval

cpi.annual = window(cpi.ts, start = c(1959,1), end = c(2010, 12), frequency = 1)

Convert nominal prices of metal to real price

• nominal price is the present price of metals
• But here, inflation has occurred twice in the selected time range.
• when we take the inflation in to account the prices may differ drastically.
• example: $100 at present is less value compared to $100 may be 50 years before.
• This is because inflation has reduced the monetary value.
• To accurately predict and see what happened in the data, we must convert metal prices to real prices.

 tungsten_real_price=(tungsten/cpi.annual)*100

plot the nominal prices and index:

• This step calculates how nominal price and index were there for the selected time interval

autoplot(tungsten)+autolayer(cpi.annual)

plot the nominal price and real price

• This step clarifies the price difference(nominal price vs real price) occurred because of inflation.

autoplot(tungsten)+autolayer(tungsten_real_price)

Plot the metal’s real price

• This helps us to see the patterns in the overall time interval.

autoplot(tungsten_real_price)

Interactively Visualizing the data.

• The above plot gives the patterns on the entire dataset.
• But, our bet is between 1980 and 1990.
• To zoom-in and see the patterns we need an interactive tool.
• dygraph from tidyverse package provides that interactivity to zoom-in and see the patterns.

 dygraphs::dygraph(tungsten_real_price, main = "Tungsten prices prediction", 
                              ylab = "Tungsten Price") %>%
                dyOptions(
                    fillGraph=TRUE, 
                    drawGrid = FALSE, 
                    colors="darkred") %>%
                dyRangeSelector() %>%
                dyCrosshair(direction = "vertical") %>%
                dyHighlight(
                    highlightCircleSize = 5, 
                    highlightSeriesBackgroundAlpha = 0.5, 
                    hideOnMouseOut = FALSE
                ) 

Conclusion:

• From the viz, it is clear that metal prices are going down from 1980 to 1990.
• Technology helped metal prices to not go high.
• paul ehlirch lost the bet to julian simon.
• As the bet is for five metals, we still have to include four more metals before coming to a final conclusion.