2Raw Dataset and Your ObjectivesFor the purposes of getting your hands dirty with real-world heterogeneous data, you'll be working with a dataset thatsubsamples 8 years of historical hourly price information for a smallish collection of financial instruments includingExchanging the US dollar for the currency of a variety of other countries within a major foreign exchange tradingplatform.A number of exchange traded funds that track the major indices in the associated countries, and an ETF thatattempts to track the change in the spot price of oil (light sweet crude delivered to Cushing, Oklahoma).Additionally, a few major indices such as the S& P 500 index, the VIX volatility index, and an index tracking theyield on United States treasuries are included. For each of these, you have an open (the price that started the hour),a close (the price the hour ended at), a high (the highest price during the hour), and a low (the lowest price duringthe hour). The instruments involved have heterogeneous trading hours, so when a market is closed, the data from themost recent hour has been filled in. There are also missing values indicated by NaNs because of heterogenous datesamong the securities when trading became available. Finally, there is a date attribute giving the date and time of thehour in question.The simple prediction task we will look at is classifying the movement of the exchange rate between the US dollarand the Canadian dollar in the next hour based on the previous hours information for these securities. Precisely, themodel will predict which of a series of discrete "bins" the fractional difference between the next hour's high and thepresent hour's close will lie in.Please realize that there are ton of other relevant quantities that a economics course would indicate as beingreasonably useful for predicting exchange rates such as this one that are not included among the features, for instance(1) sovereign debt yields in the foreign countries, (2) some measure of corporate debt yield, (3) prices of related swapsand derivatives. Do not mistake this for a proper economics assignment - the point of this assignment is just to playwith some real measured data from markets and practice operations within the tensorflow library, not to imbue anover-confidence of understanding how foreign exchange markets work.More precisely, the dataset is meant to be an example of the type of heterogenous data you may encounter andwant to work with within tensorflow and build a simple keras model for. We'll practice converting it into the tfrecordformat, then reloading it, as well as handling the different types of features we'll create from it with appropriate keraspreprocessing layers.The dataset is shared with you as a pickleappml-assignment1-dataset-v2.pklof a dictionary containing two pandasdataframes, which you can read with the pandas' readpickle method. You'll read the pickle then extract the variablesXandyfrom the python dictionary it returns. A given row of X gives the present value of the prices mentioned above,and the same row of y gives the next hour's Canadian dollar exchange rate hig