Принятие решений в бизнесе на основе данных github

Проекты курса Яндекс.Практикум «Аналитик данных»

Содержание

Yandex.Practicum_Data_Science

Certificate : https://drive.google.com/file/d/1Nl63mM8kTpNoc7LFLbsNip1_ivb_DKVv/view?usp=sharing

This repository is mainly for projects I have done under Yandex.Practicum_Data_Science.

Yandex.Practicum_Data_Science online programm helps me to achive new skill and knowledge and prepare me for a career as a data scientist. It helps to learn how to preprocess, organize and clean data, uncover pattern, insights and anomalies, visualize and repsresent the meaningful conclusions, make statistical research on the data, to learn different machine models and neural networks depending on the task and type of given data.

Tips: for data science project with python, I would recommend to install pandas, numpy, scipy, scikit learn, matplotlib, seaborn, tensorflow. These are the most basic libraries.

Part 1 — Data Preprocessing

Subjects covered:

• different methods of dealing with gaps and anomaly values (artifacts);
• changing the data type;
• different methods of determining and dealing with duplicates;
• data categorization.

Project 1 — Research on the reliability of borrowers

Bank credit department want to analyze how family status, number of kids and other factors. Given data — statitics abour bank clients paying capacity.

Part 2 — Data Analyze

Subjets covered:

• data visualisation(plots, histogramms, boxplots, subplots and etc.);
• data slices and filtering (query, logic states, slices);
• different methods of joining data frames and data within;

Project 2 — Research of apartment sale advertisements in Saint-Petersburg

The goal of this project is to establish the correct parameters of real estate for sale. Clean the data from anomaly values and scammers data. Futher these establishment will help to determine the market values of real estate.

Part 3 — Statistic Analyze

Subjets covered:

• choose optimal metrics for data statistic description;
• evaluate discrete and continuous values using different types of histograms;
• make conclusions about data on statistical indicators;
• get acquainted with the basics probability theory;
• determine the type of distribution (normal and binominal);
• learn how to build and test hypothesis.

Project 3 — Determinition of a promising tariff for a telecom company

The goal is to make a data analyze and make a decision: which tariff is best for clients.

Project 4 — Determinition of a succesfull game for sale

The goal is to make a data analyze and to identify the patterns that determine the success of the game for sale and popularity.

Part 4 — Intro to Machine Learning

Subjects covered:

• understand the machine learning ideas and theory;
• mechine learning for regression and classification;
• scikit-learn library;
• machine learning metrics;
• research different machine learning models and determine the best for the task

Project 5 — Machine_Learning_Classification_for_telecom_company_clients

Analyze the data with clients behaivor from project 3. Need to choose the best machine learning model for classification and choose the best model depending on the clients listed in the database. Test the final model with best hyperparameters on the test samples and compare with dummy model for adecvate predictions of the chosen model. For metrics — accuracy score.

Part 5 — Machine Learning

Subjects covered:

• classification metrics (accuracy, f1, AUC-ROC);
• classification model with classes disbalance;
• regression metrics (MAE and R^2);

Project 6 — Prediction the bank client behaivor (stay or leave)

Analyze the client data to predict the behaivor of each client will it stay or leave the bank service. Using F1 and AUC-ROC metrics train and choose the best model for this goal.

Part 6 — Machine Learning in business

Subjects covered:

• determine the correct metrics for business;
• lifetime of a data model from scratch to deployment and business usage;
• real business cases

Project 7 — Predict a location for an oil well

In the region characterisits of different oil wells have been collected (oil quality and oil volume for each well). Depending on this data oil volume and quality is predicted for new wells. It is more efficienty to use a ML to predict then to test each oil well manually. After prediction the region with the maximum total profit is selected according to the oil wells predicted by model.

Project 8 — Predict recovery factor of gold from gold ore.

When the mined ore undergoes primary processing, a crushed mixture is obtained. It is sent to flotation (beneficiation) and two-stage purification. After this technological process the recovery factor of gold must be predicted.

Part 7 — Gradient Boosting

Subjects covered:

• algorithm of gradient boosting;
• gradient boosting using python;
• tunning the gradient boosting;
• libraries for gradient boosting;

Project 9 — Car cost prediction

Service for the sale of used cars is developing an application to attract new customers. In it, you can quickly find out the market value of your car. Using Historical data: technical characteristics, equipment and prices of cars need to build a model to determine the cost.

Part 8 — Time Rows

Subjects covered:

• learn to use machine learning with time rows;
• learn to analyze trends and seasons in time rows;
• learn to create features from datetime data;

Project 10 — Taxi order prediction

Relying on the past data of taxi orders (datetime data). The goal is to predict the amount of taxi orders.

Part 9 — Machine Learning for text

Subjects covered:

• text vectorization;
• TF-IDF;
• text sentiment analysis;
• Embeddings, Word2Vec, BERT and etc.

Project 11 — Determine toxic commentaries in a shop-service

The goal is to train a classification model for determining toxic comments in shop service app.

Part 10 — Data Extraction (SQL practice)

Subjects covered:

• what is data bases (SQL);
• SQL practice;
• PySpark;

Project 12 — Data Analyze the SQL database

THe goal is to analyze what are the preferences of airline clients are. Used the given SQL database to collect the needed data.

Part 11 — Neural Networks

Subjects covered:

• fully connected neural networks
• convolition neural networks
• Keras, TensorFlow
• LeNet, Adam arhitecture
• Augemntation types

Project 13 — Shop customers age prediction

The goal is to train a neural network with photo set of customers to determine the age of customers in future.

Part 12 — Unsupervised Learning

• searching anomalies and clusterization
• k-means, knn-method, isolation forest

Project 14 — Temperature prediction for steel production

The goal is to analyze the data given from different datasets. understand the process of steel production, and train the model which will determine temperature of steel in the end of technological process.