What is Problem Formulation in Machine Learning and Top 4 examples of Problem Formulation in Machine Learning?

Machine Learning (ML) is a field of Artificial Intelligence (AI) that enables computers to learn from data, without being explicitly programmed. Machine learning algorithms build models based on sample data, known as “training data”, in order to make predictions or decisions, rather than following rules written by humans. Machine learning is closely related to and often overlaps with computational statistics; a discipline that also focuses on prediction-making through the use of computers. Machine learning can be applied in a wide variety of domains, such as medical diagnosis, stock trading, robot control, manufacturing and more.

The process of machine learning consists of several steps: first, data is collected; then, a model is selected or created; finally, the model is trained on the collected data and then applied to new data. This process is often referred to as the “machine learning pipeline”. Problem formulation is the second step in this pipeline and it consists of selecting or creating a suitable model for the task at hand and determining how to represent the collected data so that it can be used by the selected model. In other words, problem formulation is the process of taking a real-world problem and translating it into a format that can be solved by a machine learning algorithm.

There are many different types of machine learning problems, such as classification, regression, prediction and so on. The choice of which type of problem to formulate depends on the nature of the task at hand and the type of data available. For example, if we want to build a system that can automatically detect fraudulent credit card transactions, we would formulate a classification problem. On the other hand, if our goal is to predict the sale price of houses given information about their size, location and age, we would formulate a regression problem. In general, it is best to start with a simple problem formulation and then move on to more complex ones if needed.

Some common examples of problem formulations in machine learning are:
– Classification: given an input data point (e.g., an image), predict its category label (e.g., dog vs cat).
– Regression: given an input data point (e.g., size and location of a house), predict a continuous output value (e.g., sale price).
– Prediction: given an input sequence (e.g., a series of past stock prices), predict the next value in the sequence (e.g., future stock price).
– Anomaly detection: given an input data point (e.g., transaction details), decide whether it is normal or anomalous (i.e., fraudulent).
– Recommendation: given information about users (e.g., age and gender) and items (e.g., books and movies), recommend items to users (e.g., suggest books for someone who likes romance novels).
– Optimization: given a set of constraints (e.g., budget) and objectives (e.g., maximize profit), find the best solution (e.g., product mix).

ML For Dummies on iOs

ML PRO without ADS on iOs [No Ads]

ML PRO without ADS on Windows [No Ads]

ML PRO For Web/Android on Amazon [No Ads]

Problem Formulation: What this pipeline phase entails and why it’s important

The problem formulation phase of the ML Pipeline is critical, and it’s where everything begins. Typically, this phase is kicked off with a question of some kind. Examples of these kinds of questions include: Could cars really drive themselves?  What additional product should we offer someone as they checkout? How much storage will clients need from a data center at a given time?

The problem formulation phase starts by seeing a problem and thinking “what question, if I could answer it, would provide the most value to my business?” If I knew the next product a customer was going to buy, is that most valuable? If I knew what was going to be popular over the holidays, is that most valuable? If I better understood who my customers are, is that most valuable?

However, some problems are not so obvious. When sales drop, new competitors emerge, or there’s a big change to a company/team/org, it can be easy to say, “I see the problem!” But sometimes the problem isn’t so clear. Consider self-driving cars. How many people think to themselves, “driving cars is a huge problem”? Probably not many. In fact, there isn’t a problem in the traditional sense of the word but there is an opportunity. Creating self-driving cars is a huge opportunity. That doesn’t mean there isn’t a problem or challenge connected to that opportunity. How do you design a self-driving system? What data would you look at to inform the decisions you make? Will people purchase self-driving cars?

Part of the problem formulation phase includes seeing where there are opportunities to use machine learning.

In the following practice examples, you are presented with four different business scenarios. For each scenario, consider the following questions:

1. Is machine learning appropriate for this problem, and why or why not?
2. What is the ML problem if there is one, and what would a success metric look like?
3. What kind of ML problem is this?
4. Is the data appropriate?’

The solutions given in this article are one of the many ways you can formulate a business problem.

I)  Amazon recently began advertising to its customers when they visit the company website. The Director in charge of the initiative wants the advertisements to be as tailored to the customer as possible. You will have access to all the data from the retail webpage, as well as all the customer data.

1. ML is appropriate because of the scale, variety and speed required. There are potentially thousands of ads and millions of customers that need to be served customized ads immediately as they arrive to the site.
2. The problem is ads that are not useful to customers are a wasted opportunity and a nuisance to customers, yet not serving ads at all is a wasted opportunity. So how does Amazon serve the most relevant advertisements to its retail customers?
   1. Success would be the purchase of a product that was advertised.
3. This is a supervised learning problem because we have a labeled data point, our success metric, which is the purchase of a product.
4. This data is appropriate because it is both the retail webpage data as well as the customer data.

II) You’re a Senior Business Analyst at a social media company that focuses on streaming. Streamers use a combination of hashtags and predefined categories to be discoverable by your platform’s consumers. You ran an analysis on unique streamer counts by hashtags and categories over the last month and found that out of tens of thousands of streamers, almost all use only 40 hashtags and 10 categories despite innumerable hashtags and hundreds of categories. You presume the predefined categories don’t represent all the possibilities very well, and that streamers are simply picking the closest fit. You figure there are likely many categories and groupings of streamers that are not accounted for. So you collect a dataset that consists of all streamer profile descriptions (all text), all the historical chat information for each streamer, and all their videos that have been streamed.

1. ML is appropriate because of the scale and variability.
2. The problem is the content of streamers is not being represented by the existing categories. Success would be naturally grouping the streamers into categories based on content and seeing if those align with the hashtags and categories that are being commonly used.  If they do not, then the streamers are not being well represented and you can use these groupings to create new categories.
3. There isn’t a specific outcome variable. There’s no target or label. So this is an unsupervised problem.
4. The data is appropriate.

III) You’re a headphone manufacturer who sells directly to big and small electronic stores. As an attempt to increase competitive pricing, Store 1 and Store 2 decided to put together the pricing details for all headphone manufacturers and their products (about 350 products) and conduct daily releases of the data. You will have all the specs from each manufacturer and their product’s pricing. Your sales have recently been dropping so your first concern is whether there are competing products that are priced lower than your flagship product.

1. ML is probably not necessary for this. You can just search the dataset to see which headphones are priced lower than the flagship, then compare their features and build quality.

IV) You’re a Senior Product Manager at a leading ridesharing company. You did some market research, collected customer feedback, and discovered that both customers and drivers are not happy with an app feature. This feature allows customers to place a pin exactly where they want to be picked up. The customers say drivers rarely stop at the pin location. Drivers say customers most often put the pin in a place they can’t stop. Your company has a relationship with the most used maps app for the driver’s navigation so you leverage this existing relationship to get direct, backend access to their data. This includes latitude and longitude, visual photos of each lat/long, traffic delay details, and regulation data if available (ie- No Parking zones, 3 minute parking zones, fire hydrants, etc.).

1. ML is appropriate because of the scale and automation involved. It’s not feasible to drive everywhere and write down all the places that are ok for pickup. However, maybe we can predict whether a location is ok for pickup.
2. The problem is drivers and customers are having poor experiences connecting for pickup, which is pushing customers away from the platform.
   1. Success would be properly identifying appropriate pickup locations so they can be integrated into the feature.
3. This is a supervised learning problem even though there aren’t any labels, yet. Someone will have to go through a sample of the data to label where there are ok places to park and not park, giving the algorithms some target information.
4. The data is appropriate once a sample of the dataset has been labeled. There may be some other data that could be included too. What about asking UPS for driver stop information? Where do they stop?

In conclusion, problem formulation is an important step in the machine learning pipeline that should not be overlooked or underestimated. It can make or break a machine learning project; therefore, it is important to take care when formulating machine learning problems.”

Step by Step Solution to a Machine Learning Problem – Feature Engineering

Feature Engineering is the act of reshaping and curating existing data to make patters more apparent. This process makes the data easier for an ML model to understand. Using knowledge of the data, features are engineered and  tuned to make ML algorithms work more efficiently.

Invest in your future today by enrolling in this Azure Fundamentals - Pass the Azure Fundamentals Exam with Ease: Master the AZ-900 Certification with the Comprehensive Exam Preparation Guide!

For this problem, imagine a scenario where you are running a real estate brokerage and you want to predict the selling price of a house. Using a specific county dataset and simple information (like the location, total square footage, and number of bedrooms), let’s practice training a baseline model, conducting feature engineering, and tuning a model to make a prediction.

First, load the dataset and take a look at its basic properties.

# Load the dataset
import pandas as pd
import boto3

df = pd.read_csv(“xxxxx_data_2.csv”)
df.head()

Output:

This dataset has 21 columns:

• id – Unique id number
• date – Date of the house sale
• price – Price the house sold for
• bedrooms – Number of bedrooms
• bathrooms – Number of bathrooms
• sqft_living – Number of square feet of the living space
• sqft_lot – Number of square feet of the lot
• floors – Number of floors in the house
• waterfront – Whether the home is on the waterfront
• view – Number of lot sides with a view
• condition – Condition of the house
• grade – Classification by construction quality
• sqft_above – Number of square feet above ground
• sqft_basement – Number of square feet below ground
• yr_built – Year built
• yr_renovated – Year renovated
• zipcode – ZIP code
• lat – Latitude
• long – Longitude
• sqft_living15 – Number of square feet of living space in 2015 (can differ from sqft_living in the case of recent renovations)
• sqrt_lot15 – Nnumber of square feet of lot space in 2015 (can differ from sqft_lot in the case of recent renovations)

This dataset is rich and provides a fantastic playground for the exploration of feature engineering. This exercise will focus on a small number of columns. If you are interested, you could return to this dataset later to practice feature engineering on the remaining columns.

A baseline model

Now, let’s  train a baseline model.

People often look at square footage first when evaluating a home. We will do the same in the oflorur model and ask how well can the cost of the house be approximated based on this number alone. We will train a simple linear learner model (documentation). We will compare to this after finishing the feature engineering.

import sagemaker
import numpy as np
from sklearn.model_selection import train_test_split
import time

t1 = time.time()

# Split training, validation, and test
ys = np.array(df[‘price’]).astype(“float32”)
xs = np.array(df[‘sqft_living’]).astype(“float32”).reshape(-1,1)

np.random.seed(8675309)
train_features, test_features, train_labels, test_labels = train_test_split(xs, ys, test_size=0.2)
val_features, test_features, val_labels, test_labels = train_test_split(test_features, test_labels, test_size=0.5)

# Train model
linear_model = sagemaker.LinearLearner(role=sagemaker.get_execution_role(),
instance_count=1,
instance_type=’ml.m4.xlarge’,
predictor_type=’regressor’)

train_records = linear_model.record_set(train_features, train_labels, channel=’train’)
val_records = linear_model.record_set(val_features, val_labels, channel=’validation’)
test_records = linear_model.record_set(test_features, test_labels, channel=’test’)

linear_model.fit([train_records, val_records, test_records], logs=False)

sagemaker.analytics.TrainingJobAnalytics(linear_model._current_job_name, metric_names = [‘test:mse’, ‘test:absolute_loss’]).dataframe()

If you examine the quality metrics, you will see that the absolute loss is about $175,000.00. This tells us that the model is able to predict within an average of $175k of the true price. For a model based upon a single variable, this is not bad. Let’s try to do some feature engineering to improve on it.

Throughout the following work, we will constantly be adding to a dataframe called encoded. You will start by populating encoded with just the square footage you used previously.

encoded = df[[‘sqft_living’]].copy()

Categorical variables

Let’s start by including some categorical variables, beginning with simple binary variables.

The dataset has the waterfront feature, which is a binary variable. We should change the encoding from 'Y' and 'N' to 1 and 0. This can be done using the map function (documentation) provided by Pandas. It expects either a function to apply to that column or a dictionary to look up the correct transformation.

Binary categorical

Let’s write code to transform the waterfront variable into binary values. The skeleton has been provided below.

encoded[‘waterfront’] = df[‘waterfront’].map({‘Y’:1, ‘N’:0})

You can also encode many class categorical variables. Look at column condition, which gives a score of the quality of the house. Looking into the data source shows that the condition can be thought of as an ordinal categorical variable, so it makes sense to encode it with the order.

Ordinal categorical

Using the same method as in question 1, encode the ordinal categorical variable condition into the numerical range of 1 through 5.

encoded[‘condition’] = df[‘condition’].map({‘Poor’:1, ‘Fair’:2, ‘Average’:3, ‘Good’:4, ‘Very Good’:5})

A slightly more complex categorical variable is ZIP code. If you have worked with geospatial data, you may know that the full ZIP code is often too fine-grained to use as a feature on its own. However, there are only 7070 unique ZIP codes in this dataset, so we may use them.

However, we do not want to use unencoded ZIP codes. There is no reason that a larger ZIP code should correspond to a higher or lower price, but it is likely that particular ZIP codes would. This is the perfect case to perform one-hot encoding. You can use the get_dummies function (documentation) from Pandas to do this.

Nominal categorical

Using the Pandas get_dummies function,  add columns to one-hot encode the ZIP code and add it to the dataset.

encoded = pd.concat([encoded, pd.get_dummies(df[‘zipcode’])], axis=1)

In this way, you may freely encode whatever categorical variables you wish. Be aware that for categorical variables with many categories, something will need to be done to reduce the number of columns created.

One additional technique, which is simple but can be highly successful, involves turning the ZIP code into a single numerical column by creating a single feature that is the average price of a home in that ZIP code. This is called target encoding.

To do this, use groupby (documentation) and mean (documentation) to first group the rows of the DataFrame by ZIP code and then take the mean of each group. The resulting object can be mapped over the ZIP code column to encode the feature.

Nominal categorical II

Complete the following code snippet to provide a target encoding for the ZIP code.

means = df.groupby(‘zipcode’)[‘price’].mean()
encoded[‘zip_mean’] = df[‘zipcode’].map(means)

Normally, you only either one-hot encode or target encode. For this exercise, leave both in. In practice, you should try both, see which one performs better on a validation set, and then use that method.

Scaling

Take a look at the dataset. Print a summary of the encoded dataset using describe (documentation).

encoded.describe()

One column ranges from 290290 to 1354013540 (sqft_living), another column ranges from 11 to 55 (condition), 7171 columns are all either 00 or 11 (one-hot encoded ZIP code), and then the final column ranges from a few hundred thousand to a couple million (zip_mean).

In a linear model, these will not be on equal footing. The sqft_living column will be approximately 1300013000 times easier for the model to find a pattern in than the other columns. To solve this, you often want to scale features to a standardized range. In this case, you will scale sqft_living to lie within 00 and 11.

Feature scaling

Fill in the code skeleton below to scale the column of the DataFrame to be between 00 and 11.

sqft_min = encoded[‘sqft_living’].min()
sqft_max = encoded[‘sqft_living’].max()
encoded[‘sqft_living’] = encoded[‘sqft_living’].map(lambda x : (x-sqft_min)/(sqft_max – sqft_min))

cond_min = encoded[‘condition’].min()
cond_max = encoded[‘condition’].max()
encoded[‘condition’] = encoded[‘condition’].map(lambda x : (x-cond_min)/(cond_max – cond_min))]

Read more here….

Amazon Reviews Solution

Predicting Credit Card Fraud Solution

Predicting Airplane Delays Solution

Data Processing for Machine Learning Example

Model Training and Evaluation Examples

Targeting Direct Marketing Solution