<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Mekha on Medium]]></title>
        <description><![CDATA[Stories by Mekha on Medium]]></description>
        <link>https://medium.com/@mekhasabu.official?source=rss-45dc40be06e0------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*hEevPCASH2H41iN9TG4vNg.png</url>
            <title>Stories by Mekha on Medium</title>
            <link>https://medium.com/@mekhasabu.official?source=rss-45dc40be06e0------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 02:28:28 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@mekhasabu.official/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Feature Engineering in Depth: Elevating Machine Learning Performance]]></title>
            <link>https://medium.com/@mekhasabu.official/feature-engineering-in-depth-elevating-machine-learning-performance-9ccb7fcf8d97?source=rss-45dc40be06e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/9ccb7fcf8d97</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[feature-engineering]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[towards-data-science]]></category>
            <dc:creator><![CDATA[Mekha]]></dc:creator>
            <pubDate>Thu, 23 Jan 2025 16:21:23 GMT</pubDate>
            <atom:updated>2025-01-23T16:21:23.236Z</atom:updated>
            <content:encoded><![CDATA[<p>Transforming raw data into model-ready features for optimal results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LruTu3OXvKmC106COEsr4A.jpeg" /></figure><p><strong>Introduction to Feature Engineering</strong></p><p>Feature engineering is the process of transforming raw data into features that are optimized for machine learning models. It involves selecting, extracting, and modifying the most relevant features from the available data to build more accurate and efficient models.</p><p>The performance of machine learning models depends significantly on the quality of the features used to train them. Feature engineering employs a range of techniques to create new features by combining or transforming existing ones. These techniques help uncover important patterns and relationships within the data, enabling the model to learn and generalize more effectively.</p><p><em>Coming up with features is difficult, time-consuming, requires expert knowledge. “Applied machine learning” is basically feature engineering.</em></p><p><em>-Andrew Ng</em></p><p>In this blog, we will explore key aspects of feature engineering, including handling missing data, encoding variables, transforming data, discretization, and managing outliers.</p><p><strong>Missing Data</strong></p><p>Missing values occur in a dataset when certain information is not recorded for a variable. Understanding the mechanism of missing data is crucial for handling it effectively. There are three primary types of missing data mechanisms:</p><ol><li><strong><em>Missing Completely at Random (MCAR):</em></strong><br>Data is considered MCAR if the probability of a value being missing is unrelated to both the observed and missing data. In such cases, the missing values are distributed randomly throughout the dataset. For example, in a survey, if missing responses occur randomly and are unrelated to any other variables, the data is MCAR.</li><li><strong><em>Missing at Random (MAR):</em></strong><br>Data is MAR when the probability of a value being missing depends on the observed data but not the missing data itself. For instance, if older participants in a study are less likely to report their income, but the missingness is unrelated to their actual income values, the data is MAR.</li><li><strong><em>Missing Not at Random (MNAR):</em></strong><br>Data is MNAR when the probability of missing values depends on the value of the missing data itself. For example, if employees with lower job satisfaction are less likely to report their income, the data is MNAR because the missingness is related to an unobserved factor — job satisfaction.</li></ol><p><strong>Handling the Missing Data</strong></p><p><strong><em>Complete Case Analysis</em></strong></p><p>Complete Case Analysis (CCA), also known as list-wise deletion, involves analyzing only those observations in the dataset that have no missing values in any of the variables. In this method, all observations with missing values are removed, making it suitable when the dataset has very few missing observations.</p><p>While CCA is straightforward and can be applied to both categorical and numerical variables, it has significant limitations. If multiple variables have missing data or if some variables have a high proportion of missing observations, CCA can lead to the removal of a large portion of the dataset, which is undesirable.</p><p>In practice, CCA may be acceptable only when the amount of missing data is negligible. However, in many real-world datasets where missing data is common, CCA is rarely a viable option.</p><p>Now lets take an example of titanic dataset,</p><pre>import seaborn as sns<br><br># Load Titanic dataset<br>df = sns.load_dataset(&#39;titanic&#39;)<br>df1=df.copy()<br></pre><pre># Check missing values<br>df.isnull().sum()</pre><pre>survived         0<br>pclass           0<br>sex              0<br>age            177<br>sibsp            0<br>parch            0<br>fare             0<br>embarked         2<br>class            0<br>who              0<br>adult_male       0<br>deck           688<br>embark_town      2<br>alive            0<br>alone            0<br>dtype: int64</pre><pre># Delete the rows or data point to handle missing values<br>df.dropna().shape</pre><pre>(182, 15)</pre><pre>## Column wise deletion<br>df.dropna(axis=1)</pre><p>In this scenario, where missing values are deleted, there is a significant risk of losing valuable information. For datasets with numerous variables containing missing data, Complete Case Analysis (CCA) is often not a feasible option, as it results in a reduced dataset with only complete observations. However, if only a subset of variables from the dataset is being used, we can evaluate each variable individually to decide whether to discard missing values or replace them using alternative methods.</p><p><strong><em>Imputation of Missing Values</em></strong></p><ol><li><em>Mean Value Imputation</em></li></ol><p>Mean imputation involves replacing missing values (NA) in a dataset with the mean of the variable. This method is commonly used when the data follows a <strong>Gaussian (normal) distribution</strong> .</p><p>Mean imputation operates under the assumption that the data is <strong>Missing Completely at Random (MCAR)</strong>. This means that the missing values in the dataset are not dependent on the observed or unobserved data, and therefore, replacing missing values with the mean will not introduce bias.</p><p>If this assumption holds, it is reasonable to replace the missing values with the most frequent occurrence of the variable. For Gaussian-distributed variables, this most frequent occurrence is the <strong>mean</strong>.</p><pre># Impute missing values in &#39;Age&#39; using the mean<br>mean_age = df1[&#39;Age&#39;].mean()<br>df1[&#39;Age_mean&#39;] = df1[&#39;Age&#39;].fillna(mean_age)<br><br># Display the first few rows<br>print(df1[[&#39;Age&#39;, &#39;Age_mean&#39;]].head())</pre><p>2.<em>Median Value Imputation</em></p><p>Median imputation is robust to outliers and replaces missing values with the median of the column.</p><pre>df[&#39;age_median&#39;]=df[&#39;age&#39;].fillna(df[&#39;age&#39;].median())</pre><p><em>3.Mode Value Imputation</em></p><p>Mode imputation is suitable for categorical columns and replaces missing values with the most frequent value in the column.</p><pre># Impute missing values in &#39;Embarked&#39; using the mode<br>mode_embarked = df[&#39;Embarked&#39;].mode()[0]<br>df[&#39;Embarked_mode&#39;] = df[&#39;Embarked&#39;].fillna(mode_embarked)</pre><p><em>4. Random Sampling Imputation</em></p><p>Random sampling replaces missing values with a random value sampled from the non-missing values of the column.</p><pre># Function to perform random sampling imputation<br>def random_imputation(column):<br>    missing = column.isnull()<br>    num_missing = missing.sum()<br>    sampled_values = column.dropna().sample(num_missing, random_state=42)<br>    sampled_values.index = column[missing].index<br>    column[missing] = sampled_values<br>    return column<br><br># Impute missing values in &#39;Age&#39; using random sampling<br>df[&#39;Age_random&#39;] = random_imputation(df[&#39;Age&#39;])<br><br># Display the first few rows<br>print(df[[&#39;Age&#39;, &#39;Age_random&#39;]].head())</pre><h4><strong>Handling the Imbalanced Dataset</strong></h4><p>Imbalanced datasets occur when the target variable has an unequal distribution of classes. For example, in a classification problem where 90% of the observations belong to one class and only 10% belong to the other, the classifier might become biased towards the majority class, leading to poor performance on the minority class.</p><p>Balancing a dataset ensures that the classifier is exposed to sufficient examples from both classes during training. This improves the model’s ability to generalize and perform well on both classes.</p><p><strong>Methods to Handle Imbalanced Datasets</strong></p><p><strong><em>1.Upsampling and Downsampling</em></strong></p><p>Upsampling involves increasing the number of samples in the minority class by duplicating existing samples or generating synthetic samples. This ensures the minority class has a similar number of observations as the majority class.</p><p>Downsampling involves reducing the number of samples in the majority class by randomly selecting a subset. This ensures the class distribution is balanced by reducing the size of the majority class.</p><p>Let’s create an imbalanced dataset by focusing on the &#39;Survived&#39; column and applying upsampling and downsampling to balance it.</p><pre># Creating an imbalanced dataset<br>df_imbalanced = df[df[&#39;Survived&#39;].isin([0, 1])]<br>print(&quot;Original Target Distribution:&quot;)<br>print(df_imbalanced[&#39;Survived&#39;].value_counts())</pre><pre># Separate majority and minority classes<br>majority = df_imbalanced[df_imbalanced[&#39;Survived&#39;] == 0]<br>minority = df_imbalanced[df_imbalanced[&#39;Survived&#39;] == 1]<br><br># Upsample minority class<br>minority_upsampled = resample(minority,<br>                              replace=True,<br>                              n_samples=len(majority),<br>                              random_state=42)<br><br># Combine majority and upsampled minority<br>df_upsampled = pd.concat([majority, minority_upsampled])<br><br>print(&quot;\nAfter Upsampling:&quot;)<br>print(df_upsampled[&#39;Survived&#39;].value_counts())</pre><pre># Downsample majority class<br>majority_downsampled = resample(majority,<br>                                replace=False,<br>                                n_samples=len(minority),<br>                                random_state=42)<br><br># Combine minority and downsampled majority<br>df_downsampled = pd.concat([majority_downsampled, minority])<br><br>print(&quot;\nAfter Downsampling:&quot;)<br>print(df_downsampled[&#39;Survived&#39;].value_counts())</pre><p><strong><em>2.SMOTE (Synthetic Minority Oversampling Technique)</em></strong></p><p><strong>SMOTE</strong> is a powerful technique in machine learning used to address imbalanced datasets. It works by generating synthetic instances for the minority class, rather than simply duplicating existing ones. These synthetic samples are created by interpolating between existing instances of the minority class.</p><p>How SMOTE works</p><ul><li>SMOTE identifies the <strong>k-nearest neighbors</strong> of each minority class instance.</li><li>It selects one of these neighbors at random.</li><li>A synthetic sample is generated by interpolating between the original instance and the selected neighbor.</li></ul><p>We can create a sample imbalanced dataset to study SMOTE:</p><pre>from collections import Counter<br>from sklearn.datasets import make_classification<br>from imblearn.over_sampling import SMOTE<br><br>X, y = make_classification(n_classes=2, class_sep=2,<br>weights=[0.1, 0.9], n_informative=3, n_redundant=1, flip_y=0,<br>n_features=20, n_clusters_per_class=1, n_samples=1000, random_state=10)<br>print(&#39;Original dataset shape %s&#39; % Counter(y))</pre><pre>Original dataset shape Counter({1: 900, 0: 100})</pre><pre>sm = SMOTE(random_state=42)<br>X_res, y_res = sm.fit_resample(X, y)<br>print(&#39;Resampled dataset shape %s&#39; % Counter(y_res))</pre><pre>Resampled dataset shape Counter({0: 900, 1: 900})</pre><h4><strong>Data Encoding</strong></h4><p>Many machine learning algorithms require numerical input. Encoding categorical variables into numerical formats ensures they can be properly utilized by these algorithms. However, the choice of encoding technique depends on the type of categorical data and the specific use case.</p><p>Categorical variables are divided into:</p><ol><li><strong>Nominal Variables</strong>: No intrinsic order (e.g., colors, cities).</li><li><strong>Ordinal Variables</strong>: Intrinsic order exists (e.g., education levels, ratings).</li></ol><p>Here’s an explanation of key encoding methods:</p><p><strong>1.One-Hot Encoding</strong></p><p>One-Hot Encoding, also known as <strong>Nominal Encoding</strong>, is a technique used to represent categorical data in a numerical format, making it more suitable for machine learning algorithms. This encoding method is commonly applied to <strong>nominal variables</strong> — categorical variables with no intrinsic order or ranking.</p><p>In One-Hot Encoding, each unique category is transformed into a binary vector where each bit corresponds to a unique category. For instance, if we have a categorical variable color with three possible values—red, green, and blue—they can be encoded as:</p><ol><li><strong>Red</strong> → [1, 0, 0]</li><li><strong>Green</strong> → [0, 1, 0]</li><li><strong>Blue</strong> → [0, 0, 1]</li></ol><p>Each vector contains only one 1 to represent the category while the rest are 0.</p><p>Below is an example of how to apply One-Hot Encoding using Python and scikit-learn.</p><pre>import pandas as pd<br>from sklearn.preprocessing import OneHotEncoder<br><br># Step 1: Create a simple dataframe<br>df = pd.DataFrame({<br>    &#39;color&#39;: [&#39;red&#39;, &#39;blue&#39;, &#39;green&#39;, &#39;green&#39;, &#39;red&#39;, &#39;blue&#39;]<br>})<br><br># Step 2: Create an instance of OneHotEncoder<br>encoder = OneHotEncoder()<br><br># Step 3: Perform fit and transform<br># Transform the &#39;color&#39; column into a one-hot encoded array<br>encoded_array = encoder.fit_transform(df[[&#39;color&#39;]]).toarray()<br><br># Step 4: Convert the encoded array into a DataFrame for better readability<br>encoded_df = pd.DataFrame(encoded_array, columns=encoder.get_feature_names_out([&#39;color&#39;]))<br><br># Combine the original and encoded data<br>final_df = pd.concat([df, encoded_df], axis=1)<br><br># Display the final DataFrame<br>print(final_df)</pre><p><strong>2.Label Encoding<br></strong>Label Encoding is a straightforward technique used to convert categorical variables into numerical data by assigning a unique integer to each category. This method is particularly useful for variables where categories are not inherently ordinal, but its simplicity can also lead to unintended consequences.</p><p>For example, if we have a categorical variable color with three categories—red, green, and blue—Label Encoding might assign:</p><ol><li><strong>Red</strong> → 0</li><li><strong>Green</strong> → 1</li><li><strong>Blue</strong> → 2</li></ol><p>Label Encoding works well for machine learning algorithms that do not rely on numerical magnitude or order, such as decision trees. However, for algorithms like linear regression, the artificial ordering could lead to misleading results, as the encoded numbers may imply relationships between categories that do not exist.</p><pre>from sklearn.preprocessing import LabelEncoder<br><br># Create an instance of LabelEncoder<br>label_encoder = LabelEncoder()<br><br># Perform fit and transform<br># Transform the &#39;color&#39; column into numerical labels<br>encoded_labels = label_encoder.fit_transform(df[&#39;color&#39;])<br><br># Add the encoded labels as a new column in the dataframe<br>df[&#39;color_encoded&#39;] = encoded_labels<br><br># Display the final DataFrame<br>print(df)</pre><p><strong>3.Ordinal Encoding</strong></p><p>Ordinal Encoding is a technique used to encode categorical data that has an inherent order or ranking. Each category is assigned a numerical value based on its position in the order, making it ideal for features like education level, income brackets, or satisfaction ratings. For example, if the education level variable has four categories—high school, college, graduate, and post-graduate—Ordinal Encoding can represent them as:</p><ol><li><strong>High school</strong> → 1</li><li><strong>College</strong> → 2</li><li><strong>Graduate</strong> → 3</li><li><strong>Post-graduate</strong> → 4</li></ol><p>This method ensures that the relationship between categories is preserved, which is particularly useful in models that rely on the ordinal nature of the data.<br>Below is an example of applying Ordinal Encoding using the OrdinalEncoder class from sklearn.preprocessing.</p><pre>import pandas as pd<br>from sklearn.preprocessing import OrdinalEncoder<br><br># Step 1: Create a simple dataframe<br>df = pd.DataFrame({<br>    &#39;education_level&#39;: [&#39;high school&#39;, &#39;college&#39;, &#39;graduate&#39;, &#39;post-graduate&#39;, &#39;high school&#39;, &#39;graduate&#39;]<br>})<br><br># Step 2: Define the order of categories<br>education_order = [&#39;high school&#39;, &#39;college&#39;, &#39;graduate&#39;, &#39;post-graduate&#39;]<br><br># Step 3: Create an instance of OrdinalEncoder with the defined category order<br>ordinal_encoder = OrdinalEncoder(categories=[education_order])<br><br># Step 4: Perform fit and transform<br># Transform the &#39;education_level&#39; column into ordinal values<br>df[&#39;education_encoded&#39;] = ordinal_encoder.fit_transform(df[[&#39;education_level&#39;]])<br><br># Display the final DataFrame<br>print(df)</pre><p><strong>4.Target Guided Ordinal Encoding<br></strong>Target Guided Ordinal Encoding is a technique that encodes categorical variables based on their statistical relationship with the target variable. This approach is particularly effective for categorical variables with a high cardinality (a large number of unique categories).</p><p>Instead of assigning arbitrary numerical values to categories, Target Guided Ordinal Encoding calculates a summary statistic (typically the <strong>mean</strong> or <strong>median</strong>) of the target variable for each category. These calculated values are then used to replace the categories in the variable.</p><p>For instance, if we have a categorical variable city and a target variable house price, we might replace each city with the mean house price for that city. This encoding establishes a monotonic relationship between the feature and the target, enhancing the predictive power of certain machine learning models.<br>Let’s demonstrate Target Guided Ordinal Encoding using a dataset with a City feature and a Sales target variable.</p><pre>import pandas as pd<br><br># Step 1: Create a sample dataset<br>data = {<br>    &#39;City&#39;: [&#39;New York&#39;, &#39;London&#39;, &#39;Paris&#39;, &#39;New York&#39;, &#39;London&#39;, &#39;Paris&#39;, &#39;New York&#39;],<br>    &#39;Sales&#39;: [200, 150, 180, 220, 170, 190, 210]<br>}<br>df = pd.DataFrame(data)<br><br># Step 2: Calculate the mean Sales for each City<br>city_sales_mean = df.groupby(&#39;City&#39;)[&#39;Sales&#39;].mean()<br><br># Step 3: Map the mean Sales values to the City column<br>df[&#39;City_encoded&#39;] = df[&#39;City&#39;].map(city_sales_mean)<br><br># Display the final DataFrame<br>print(df)</pre><h3>Discretization</h3><p>Discretization, also known as binning, is a technique used to transform continuous numerical variables into discrete categorical variables by dividing their range into intervals or “bins.” Each bin is assigned a unique label, making it easier for algorithms to identify patterns in continuous data.</p><p>This technique is commonly used in scenarios where the precise value of a variable is less important than the range it falls into. For instance, converting an age variable into categories like &quot;child,&quot; &quot;adult,&quot; and &quot;senior&quot; can simplify analysis and make data more interpretable.</p><p><strong><em>Types of Discretization</em></strong></p><p>1.<strong>Equal-Width Binning</strong>:</p><p>Divides the range of the variable into bins of equal size (width). The intervals are determined by dividing the range (maximum — minimum) into k equal parts, where k is the desired number of bins.Best suited for evenly distributed data where the range of the variable matters more than the frequency of values in each bin.If the range of a variable is 0–30 and we create 3 bins, the intervals would be [0–10), [10–20), and [20–30).</p><pre>import pandas as pd<br>import numpy as np<br><br># Step 1: Create a sample dataset<br>data = {&#39;Age&#39;: [5, 12, 17, 24, 30, 45, 60, 75]}<br>df = pd.DataFrame(data)<br><br># Step 2: Perform equal-width binning<br>df[&#39;Age_binned&#39;] = pd.cut(df[&#39;Age&#39;], bins=3, labels=[&#39;Young&#39;, &#39;Adult&#39;, &#39;Senior&#39;])<br><br># Display the final DataFrame<br>print(df)</pre><p><strong>2.Equal-Frequency Binning<br></strong>Divides the data such that each bin contains an approximately equal number of data points. The bin edges are determined based on the data distribution rather than the range.Useful when the distribution of values is uneven or skewed, as it ensures that each bin has similar representation in terms of data points.For a dataset with values [1, 2, 3, 10, 20, 30], dividing into 3 equal-frequency bins might result in bins like [1–3), [10–20), and [30+).</p><pre># Perform equal-frequency binning<br>df[&#39;Equal_Frequency_Binning&#39;] = pd.qcut(df[&#39;Age&#39;], q=3, labels=[&#39;Young&#39;, &#39;Adult&#39;, &#39;Senior&#39;])<br><br># Display the results<br>print(df)</pre><p><strong>3.Custom Binning<br></strong>Allows manual specification of bin ranges and labels based on domain knowledge or specific business needs.Ideal for scenarios where domain expertise or specific thresholds define meaningful categories (e.g., income brackets or risk levels).For example,splitting income into bins like [Low: &lt;$30,000], [Medium: $30,000–$70,000], and [High: &gt;$70,000].</p><pre># Define custom bin edges and labels<br>bins = [0, 18, 35, 60, 100]<br>labels = [&#39;Child&#39;, &#39;Young Adult&#39;, &#39;Middle Aged&#39;, &#39;Senior&#39;]<br><br># Perform custom binning<br>df[&#39;Custom_Binning&#39;] = pd.cut(df[&#39;Age&#39;], bins=bins, labels=labels)<br><br># Display the results<br>print(df)</pre><h3>Handling Outliers</h3><p>Outliers are data points that significantly deviate from the rest of the dataset. These points can occur due to measurement errors, variability in the data, or rare events. Outliers can skew statistical analyses, reduce model accuracy, and lead to misleading insights if not handled properly.<br>Methods to Detect Outliers</p><ol><li><strong>5-Number Summary</strong></li><li><strong>Boxplot Visualization</strong></li></ol><h4>1. The 5-Number Summary</h4><p>The 5-number summary is a statistical method to summarize the distribution of a dataset. It includes:</p><ul><li><strong>Minimum</strong>: The smallest value in the dataset.</li><li><strong>First Quartile (Q1)</strong>: The 25th percentile, where 25% of the data lies below this value.</li><li><strong>Median (Q2)</strong>: The middle value of the dataset, dividing it into two halves.</li><li><strong>Third Quartile (Q3)</strong>: The 75th percentile, where 75% of the data lies below this value.</li><li><strong>Maximum</strong>: The largest value in the dataset.</li></ul><p>Outliers can be identified using the <strong>Interquartile Range (IQR)</strong>:</p><ul><li>IQR=Q3−Q1</li><li>Outlier thresholds<br>1.Lower Bound: Q1−1.5×IQR<br>2.Upper Bound: Q3+1.5×IQR</li></ul><p>Calculating 5-Number Summary and Detecting Outliers:</p><pre>import numpy as np<br>import pandas as pd<br><br># Step 1: Create a dataset<br>data = [10, 12, 15, 18, 20, 22, 24, 35, 50, 100]  # Example data with potential outliers<br>df = pd.DataFrame(data, columns=[&#39;Value&#39;])<br><br># Step 2: Calculate 5-number summary<br>minimum = np.min(data)<br>maximum = np.max(data)<br>q1 = np.percentile(data, 25)<br>median = np.percentile(data, 50)<br>q3 = np.percentile(data, 75)<br>iqr = q3 - q1<br><br># Step 3: Determine outlier thresholds<br>lower_bound = q1 - 1.5 * iqr<br>upper_bound = q3 + 1.5 * iqr<br><br># Step 4: Identify outliers<br>outliers = df[(df[&#39;Value&#39;] &lt; lower_bound) | (df[&#39;Value&#39;] &gt; upper_bound)]<br><br># Print the results<br>print(f&quot;5-Number Summary:&quot;)<br>print(f&quot;Minimum: {minimum}&quot;)<br>print(f&quot;Q1: {q1}&quot;)<br>print(f&quot;Median: {median}&quot;)<br>print(f&quot;Q3: {q3}&quot;)<br>print(f&quot;Maximum: {maximum}&quot;)<br>print(f&quot;IQR: {iqr}&quot;)<br>print(f&quot;Lower Bound: {lower_bound}&quot;)<br>print(f&quot;Upper Bound: {upper_bound}&quot;)<br>print(f&quot;Outliers:\n{outliers}&quot;)</pre><h4>2. Boxplot Visualization</h4><p>A <strong>Boxplot</strong> is a graphical representation of the 5-number summary. It provides a clear way to identify potential outliers as points that fall outside the “whiskers” of the plot.</p><ul><li><strong>Box</strong>: Represents the IQR, with Q1 and Q3 as its edges.</li><li><strong>Line inside the box</strong>: The median.</li><li><strong>Whiskers</strong>: Extend from Q1 and Q3 to the smallest and largest values within 1.5 × IQR.</li><li><strong>Dots outside the whiskers</strong>: Potential outliers.</li></ul><p>Creating a Boxplot:</p><pre>import matplotlib.pyplot as plt<br><br># Step 5: Create a boxplot<br>plt.figure(figsize=(8, 5))<br>plt.boxplot(data, vert=False, patch_artist=True, boxprops=dict(facecolor=&#39;lightblue&#39;, color=&#39;black&#39;))<br>plt.title(&quot;Boxplot to Identify Outliers&quot;)<br>plt.xlabel(&quot;Value&quot;)<br>plt.show()</pre><p><strong>Conclusion<br></strong>In this blog, I’ve covered the basic yet essential aspects of feature engineering, including handling missing data, encoding categorical variables, dealing with imbalanced datasets, discretizing continuous variables, and managing outliers. These foundational techniques are critical in transforming raw data into valuable features that enhance model performance. However, feature engineering is a vast field, and there are many other advanced methods that can be explored. Techniques such as polynomial feature generation, interaction features, dimensionality reduction (like PCA), feature scaling, and domain-specific transformations are also important to master for more complex datasets and models. Additionally, for time-series data, temporal feature engineering plays a key role in extracting useful patterns. As you delve deeper into feature engineering, remember that it is an iterative process that requires creativity, domain knowledge, and an understanding of the algorithms you’re working with to unlock the full potential of your data.</p><p><strong><em>Thanks for Reading!</em></strong></p><p>Connect me on LinkedIn<br><a href="https://www.linkedin.com/in/mekha-sabu-98a488218/">LinkedIn</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9ccb7fcf8d97" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Data Analytics Can Help You Make Better Decisions in Life]]></title>
            <link>https://medium.com/@mekhasabu.official/how-data-analytics-can-help-you-make-better-decisions-in-life-e3c5ee96890f?source=rss-45dc40be06e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/e3c5ee96890f</guid>
            <category><![CDATA[illumination]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[towards-data-science]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[data-analysis]]></category>
            <dc:creator><![CDATA[Mekha]]></dc:creator>
            <pubDate>Sun, 05 Jan 2025 22:23:47 GMT</pubDate>
            <atom:updated>2025-01-05T22:23:47.773Z</atom:updated>
            <content:encoded><![CDATA[<p>I have always been fascinated by how deeply data influences our everyday lives. You don’t need to be a data nerd or a tech wizard to harness the power of data. Often, a little common sense and the ability to observe trends are all it takes to make smarter decisions. Data analytics is not confined to spreadsheets or business strategies; it’s an invisible force that guides us in countless ways every single day.</p><p>Let’s explore how we unconsciously use data in our daily lives and how embracing a more data-driven mindset can help us make better decisions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*l1vFT2U3xElJLi7UJFgATw.jpeg" /></figure><p><strong><em>Checking the Weather</em></strong></p><p>One of the simplest ways we use data every day is by checking the weather. Every time you glance at a weather app to decide what to wear, you’re leveraging data. For instance, if the forecast predicts cold and rainy weather, you’ll likely grab a jacket and boots. Without realizing it, you’re relying on predictive analytics to plan your day. Weather apps analyze historical patterns and real-time data to provide accurate forecasts, enabling you to make informed choices effortlessly.</p><p><strong><em>Planning the Route</em></strong></p><p>Ever used a navigation app to avoid traffic? Apps like Google Maps or Waze use vast amounts of data from multiple sources to suggest the fastest or most efficient route. They analyze traffic congestion in real-time, historical traffic patterns, and even road closures to ensure you get to your destination quickly. This seamless integration of data analytics into your commute saves you time and frustration.</p><p><strong><em>Choosing What to Eat</em></strong></p><p>Data also plays a crucial role when it comes to food. Whether you’re deciding which restaurant to visit or debating what to cook, you’re likely influenced by online reviews, star ratings, or even nutritional data. For example, scanning through reviews on Yelp or Google allows you to compare customer experiences and ratings to make the best choice. Here, user-generated data becomes a powerful tool for decision-making.</p><p><strong><em>Shopping Smarter</em></strong></p><p>Think about the last time you shopped online or in a store. Did you compare prices or read product reviews? Websites and apps like Amazon or PriceGrabber aggregate data from different sellers to help you find the best deals. This is data analytics at work — helping you make more informed purchasing decisions and saving you money in the process.</p><p><strong><em>Monitoring your Health</em></strong></p><p>Wearable fitness trackers like Fitbit or Apple Watch are prime examples of how data empowers personal decision-making. These devices collect and analyze data on your activity levels, sleep patterns, and heart rate. By reviewing this data, you can identify trends and make healthier lifestyle choices. For instance, if you notice poor sleep patterns, you might decide to adjust your bedtime or reduce screen time before bed.</p><h4><strong><em>Why Observing Trends Matters</em></strong></h4><p>What all these examples have in common is the ability to observe trends. You don’t need complex algorithms or advanced analytical tools to benefit from data. Simply paying attention to patterns — whether it’s recurring traffic jams, consistently high restaurant ratings, or your fitness tracker’s feedback — can make a significant difference in your daily life.</p><h4>How to Embrace a Data-Driven Mindset</h4><p>To truly benefit from data, start small:</p><ul><li><strong>Be Curious</strong>: Ask questions and look for patterns in your daily activities. For example, does your productivity peak at a certain time of day?</li><li><strong>Leverage Tools</strong>: Use accessible apps and platforms that make data simple, like weather apps, fitness trackers, or budgeting tools.</li><li><strong>Trust the Numbers</strong>: While intuition is important, data can often provide a clearer and unbiased perspective.</li></ul><h4>Tips and Tricks for Using Data Analytics in Daily Life</h4><ul><li><strong><em>Start a Daily Data Habit</em></strong><em>:</em> Choose one app or tool (like a fitness tracker or budgeting app) and spend a few minutes reviewing the data it provides each day. This will help you develop a data-driven perspective over time.</li><li><strong><em>Set Small Goals</em></strong><em>:</em> Use data to track progress in areas like fitness, savings, or productivity. For example, aim for a specific step count or spending limit and adjust based on the insights you gather.</li><li><strong><em>Ask ‘Why?’</em></strong><em>: </em>When observing trends, always question why something happens. For instance, if you’re more productive in the mornings, explore how to optimize that time further.</li><li><strong><em>Keep It Simple</em></strong><em>:</em> Don’t get overwhelmed by complex analytics. Start with easy-to-use tools and straightforward observations to integrate data into your decision-making process.</li><li><strong><em>Review and Reflect</em></strong><em>:</em> Periodically assess the data you collect. Are there patterns you hadn’t noticed? Use this reflection to make incremental changes that improve your outcomes.</li></ul><h4>Conclusion</h4><p>Data analytics isn’t just for scientists or analysts. It’s a powerful tool that everyone can use to make smarter, more informed decisions in life. From planning your outfit to managing your health, data is everywhere — helping us save time, reduce uncertainty, and improve our outcomes. By embracing a data-driven mindset, we can navigate our lives more effectively and unlock opportunities that we might otherwise overlook.</p><h4>Key Takeaway</h4><p>Start small and stay consistent. Choose one area of your life where you’d like to make better decisions — whether it’s managing your finances, improving your health, or planning your schedule. Begin observing trends and using simple data tools to guide your choices. Over time, this habit will not only make your life more efficient but also empower you to tackle bigger challenges with confidence.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e3c5ee96890f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Understanding Database Types and How to Work with Them]]></title>
            <link>https://medium.com/@mekhasabu.official/understanding-database-types-and-how-to-work-with-them-5178417aff0b?source=rss-45dc40be06e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/5178417aff0b</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[towards-data-science]]></category>
            <dc:creator><![CDATA[Mekha]]></dc:creator>
            <pubDate>Tue, 12 Nov 2024 20:46:22 GMT</pubDate>
            <atom:updated>2024-11-12T20:46:22.592Z</atom:updated>
            <content:encoded><![CDATA[<p>Explore the different types of databases, their unique strengths, and how to make the most of each for efficient data management and analysis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/980/1*Rt2bz0LHuM_Fq-tyarhtAw.jpeg" /></figure><p><strong>Introduction</strong></p><p>Databases are at the heart of almost every data-related project, from business analytics to software applications. But databases come in many types, each with unique strengths suited to different tasks. Knowing which database type to use — and understanding the basics of how they work — can make a big difference in how efficiently and effectively you can work with data.</p><p>In this post, we’ll break down the main types of databases, explain how they work, and highlight what makes each one unique. By the end, you’ll have a clear picture of the database landscape and a better idea of which type might be best for your needs.</p><p><strong>The Main Types of Databases</strong></p><p><strong><em>1.Row-Store Databases (Transactional Databases)</em></strong></p><p>Row-store databases store data in rows, where each row contains all the information about a particular item. This format is efficient for “transactional” tasks like adding, updating, or deleting records.</p><p><strong>Examples</strong>: MySQL, PostgreSQL, Microsoft SQL Server, Oracle.</p><p><strong>Strengths</strong>: Great for applications that need to frequently update or modify individual records, such as managing a list of customers or products.</p><p><strong>Limitations:</strong> When performing large-scale analysis, row-store databases can be slower because they have to read entire rows — even if you only need one or two specific fields (like age or location) from each row.</p><p><strong><em>Tip</em></strong><em>: If you’re working with a row-store database, use indexes on frequently searched fields to speed up queries.</em></p><p><em>2. </em><strong><em>Column-Store Databases (Analytical Databases)</em></strong></p><p>Column-store databases store data by columns rather than rows. This setup makes it faster to retrieve specific columns across many records, which is useful for analytical queries.</p><p><strong>Examples</strong>: Amazon Redshift, Snowflake, Google BigQuery.</p><p><strong>Strengths</strong>: Excellent for analyzing large datasets, especially when you need to run calculations on selected fields. Column-stores are known for being fast and efficient for data summarization and analysis.</p><p><strong>Limitations</strong>: Not ideal for frequent updates to individual records. Because data is stored by columns, updating a single record across multiple columns can be more time-consuming.</p><p><em>Tip: If you’re analyzing data in a column-store database, focus on querying specific columns to take full advantage of its efficiency.</em></p><p><em>3. </em><strong><em>NoSQL Databases (Non-Relational Databases)</em></strong></p><p>NoSQL databases don’t follow the strict, table-based structure of SQL databases, making them more flexible for storing unstructured or semi-structured data. There are different types of NoSQL databases, including key-value stores, document stores, and graph databases.</p><p><strong>Examples</strong>: MongoDB, Cassandra, DynamoDB.</p><p><strong>Strengths</strong>: Ideal for applications that need low-latency access to data and for handling unstructured data like documents, social media posts, or JSON data.</p><p><strong>Limitations</strong>: They aren’t optimized for querying large datasets for analysis, so they’re typically used in applications rather than for analytical reporting.</p><p><strong><em>Tip</em></strong><em>: If your organization uses NoSQL, consider integrating it with a SQL-compatible data warehouse if you need to analyze the data on a large scale.</em></p><p><strong><em>4. Data Lakes and File Storage Systems</em></strong></p><p>Data lakes are storage systems designed to hold vast amounts of raw data in its original format, such as text files or images. Commonly used for big data, data lakes store unstructured or semi-structured data at a low cost.</p><p><strong>Examples</strong>: Hadoop Distributed File System (HDFS), Amazon S3.</p><p><strong>Strengths</strong>: Cost-effective for storing massive datasets, and they offer flexibility in terms of data format.</p><p><strong>Limitations</strong>: Slower for analysis, especially without a specialized tool to process and retrieve data.</p><p><strong><em>Tip</em></strong><em>: Use data lakes for long-term data storage and consider tools like Apache Spark or SQL-on-Hadoop solutions (e.g., Hive, Presto) to query the data when needed.</em></p><h4>5. Search-Based Data Stores</h4><p>These systems are built for quickly searching and retrieving data, especially useful for machine-generated data like logs.</p><p><strong>Examples</strong>: Elasticsearch, Splunk.</p><p><strong>Strengths</strong>: Very fast for retrieving specific information from large datasets.</p><p><strong>Limitations</strong>: Not designed for complex data analysis, and their query language can differ from standard SQL.</p><p><strong><em>Tip</em></strong><em>: If you need to find specific data points within large datasets (like error logs), these systems are incredibly useful.</em></p><h4>Key Trends Shaping Databases Today</h4><p>In recent years, several trends have driven database development in new directions, providing more options for data practitioners:</p><ol><li><strong><em>Big Data and Cloud Computing</em></strong><em>: </em>With the growth of big data, cloud-based databases have become popular for their scalability and flexibility. Today, cloud databases are common for both storage and analysis, allowing companies to access powerful computing resources without the high costs of on-premises infrastructure.</li><li><strong><em>Distributed Computing:</em></strong><em> </em>Distributed computing allows data to be processed across multiple machines, making it faster and more efficient for complex tasks. This trend has enabled tools like Hadoop and Spark, which make it possible to analyze huge datasets more affordably.</li><li><strong><em>In-Memory Processing</em></strong><em>:</em> Storing data in RAM (instead of on disk) for fast retrieval has become more accessible due to lower memory costs. This technique improves the speed of analytical tasks by reducing the need to read from slower disk storage.</li></ol><p><strong>Choosing and Working with a Database Type</strong></p><p>Even if you don’t get to choose the database at your organization, understanding its type and unique capabilities will help you work more effectively. Here are some quick tips to make the most of your database:</p><ul><li><strong>For Row-Store Databases</strong>: Focus on using indexes and be mindful of how tables relate to each other. Use joins carefully and aim to retrieve only the data you need.</li><li><strong>For Column-Store Databases</strong>: When possible, filter data by specific columns to take advantage of the columnar format’s speed.</li><li><strong>For NoSQL Databases</strong>: Use them for applications that require fast data retrieval, and consider a data warehouse integration if you need SQL-based analysis.</li><li><strong>For Data Lakes</strong>: Use for cost-effective storage and consider SQL-on-Hadoop tools for querying.</li><li><strong>For Search-Based Stores</strong>: Use for fast data retrieval when you need to find specific records, especially within large log files.</li></ul><p><strong>Conclusion</strong></p><p>The world of databases is diverse and constantly evolving, but understanding the basic types and their optimal uses will help you navigate it with confidence. Whether you’re managing data transactions, performing deep analysis, or storing massive unstructured datasets, there’s a database that fits the job.</p><p>SQL remains the foundation for accessing and querying data, no matter the database type. So as you work with different databases, keep building your SQL skills — they’ll serve you well no matter where the data world goes next.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5178417aff0b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why I Chose to Learn SQL as a Data Analytics Beginner.]]></title>
            <link>https://medium.com/@mekhasabu.official/why-i-chose-to-learn-sql-as-a-data-analytics-beginner-aee8fad968db?source=rss-45dc40be06e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/aee8fad968db</guid>
            <category><![CDATA[towards-data-science]]></category>
            <dc:creator><![CDATA[Mekha]]></dc:creator>
            <pubDate>Thu, 17 Oct 2024 22:47:29 GMT</pubDate>
            <atom:updated>2024-10-17T22:47:29.837Z</atom:updated>
            <content:encoded><![CDATA[<p>Wondering where to begin your journey into data analytics? Here’s how to start!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*JmJPFjodZ0Mh9Fk_" /><figcaption>Photo by <a href="https://unsplash.com/@dengxiangs?utm_source=medium&amp;utm_medium=referral">Choong Deng Xiang</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>As someone who is starting out in the field of data analytics, I quickly realized that there is a ton of information out there.Python, R, machine learning -it’s easy to get overwhelmed by the tools people say you have to learn.But doing a bit of research and talking to the folks in the field, I kept hearing the same advice over and over again <em>start with SQL</em>.</p><p>So that’s exactly what I did. And now I want to share why I think SQL is the best place to start for anyone like me who’s just beginning their data journey.</p><p><strong><em>Why SQL? It’s Simple and Powerful</em></strong></p><p>When I first heard about SQL, I thought it was going to be another complex language like python or java. But after diving into it I realized that <em>SQL is different</em>. It’s surprisingly simple, and honestly, it feels a lot like writing plain English. The first query I ever wrote was something like this:</p><pre>SELECT name, age FROM users WHERE age &gt; 30;</pre><p>It was intuitive and made sense right away.That was a big confidence booster for me because I was able to see results instantly without feeling like I needed to spend months just learning the basics.</p><p>But don’t let SQL’s simplicity fool you — it’s also incredibly <em>powerful</em>. As I started learning more, I saw how it can handle massive amounts of data quickly. Whether you’re working with a few hundred rows or millions of rows, SQL can pull and analyze that data much faster than tools like Excel. And let’s be honest, once you start dealing with real-world data, Excel just doesn’t cut it anymore.</p><p><strong><em>SQL Opens Doors to High-Paying Jobs</em></strong></p><p>One thing that drew me to data analytics was the potential to earn a good salary while working in a field that’s in high demand. After doing some digging, I learned that <em>SQL is a must-have skill </em>for almost every data-related job out there. Whether you’re going for a role in data analysis, data engineering, or data science, knowing SQL is often listed as a core requirement.</p><p>What’s cool is that SQL is a <em>gateway skill</em>— you don’t need to be a coding expert to start using it, but it can unlock more advanced opportunities in the future. In fact, many of the people I look up to in the data world have said they got their start by learning SQL. They landed their first jobs with this skill, and later picked up other tools like Python or R to expand their abilities.</p><p>And let’s face it, the thought of eventually landing a job at a company like Google, Amazon, or any other major tech company is really exciting. From what I’ve read, SQL is a skill that’s tested in a lot of data science and analytics interviews, so mastering it can really give you a leg up when applying for those positions.</p><p><strong><em>Why SQL Beats Excel for Big Data</em></strong></p><p>If you’re like me, you’ve probably used Excel for basic data work — maybe sorting columns, making pivot tables, or doing simple calculations. And for a while, I thought that was enough. But once I started working with larger datasets, I quickly ran into issues. Excel becomes slow, and sometimes even crashes when you’re dealing with thousands of rows or complex calculations. That’s where SQL really shines.</p><p>With SQL, you can handle<em> millions of rows of data </em>effortlessly. It’s designed to deal with big data in ways that Excel simply can’t. Another thing I found super helpful is how SQL allows multiple people to work with data at the same time. With Excel, if two people try to edit the same sheet, it’s a mess. But with SQL and databases, collaboration is easy and seamless.</p><p><strong><em>The Different Flavors of SQL: What You Need to Know</em></strong></p><p>One thing I didn’t realize when I started was that there are different “flavors” of SQL — like MySQL, PostgreSQL, SQL Server, and SQLite. At first, it sounded confusing, but I learned that the core SQL syntax stays the same across all these platforms. So, if you learn SQL with one system, like MySQL (which I did), you can easily pick up another like PostgreSQL without having to start over from scratch.</p><p>That’s been really reassuring for me as a beginner. It means I’m not wasting my time learning something too specific — I’m building a foundation that will be useful no matter which type of database I end up working with in the future.</p><p><strong><em>The Path Forward: SQL is Just the Beginning</em></strong></p><p>As I keep learning SQL, I’m starting to see how it connects to the bigger picture in data analytics. SQL is just the first step, but it’s a crucial one. Once I’m comfortable with SQL, I plan to dive deeper into data visualization tools like Tableau and maybe even start learning Python. But I’m glad I started here, because everything else in data analytics seems to build on the skills SQL teaches you —<em> how to query data, analyze it, and make sense of it.</em></p><p>If you’re just starting out like me, my advice is simple:<em> start with SQL</em>. It’s beginner-friendly, powerful, and the foundation for so many data careers. Plus, it’s really rewarding to write a query and instantly see the results of your work. SQL is helping me build confidence as I step into this new world of data analytics, and I’m excited to see where it will take me next.</p><p>So, if you’re looking to kick off your data career, don’t get lost in the noise of a thousand programming languages and tools. Start with SQL, and you’ll be setting yourself up for success in the data world.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aee8fad968db" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Living in the “now”]]></title>
            <link>https://medium.com/@mekhasabu.official/living-in-the-now-852d65ff2857?source=rss-45dc40be06e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/852d65ff2857</guid>
            <dc:creator><![CDATA[Mekha]]></dc:creator>
            <pubDate>Wed, 01 May 2024 06:43:03 GMT</pubDate>
            <atom:updated>2024-05-01T06:43:03.223Z</atom:updated>
            <content:encoded><![CDATA[<p>Learn to live in the present.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*77Lb2YXufFRCOu44" /><figcaption>Photo by <a href="https://unsplash.com/@websbykaja?utm_source=medium&amp;utm_medium=referral">Kaja Kadlecova</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>“<em>Realize deeply that the present moment is all you ever have. Make the now the primary focus of your life”- Eckhart Tolle</em></p><p>Two months ago, I found myself sinking into depression and loneliness due unforeseen circumstances in my life. My mind was constantly dwelling on past regrets and future anxieties. This cycle of overthinking took a toll on my mental well-being. It felt as though I was adrift in a sea of past and future, disconnected from the vitality of present.</p><p>In this very moment, I have come to the profound realization that I possess neither the superpower to change the past nor the foresight of an astrologer to predict the future.</p><p><em>“The present moment is the only moment available to us”</em></p><p><strong>Embrace the present</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yITkwqId826HhdOE" /><figcaption>Photo by <a href="https://unsplash.com/@hadijasaidi?utm_source=medium&amp;utm_medium=referral">Hadija</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>I have read that past is the water under the bridge and and future is the distant sun on the horizon of your imagination. The most important moment is now. Learn to live in it and savor it fully.</p><p>The best path to personal fulfillment is to pursue activities that genuinely challenge you. However, it’s important to keep in mind that “happiness is a journey not a destination”. Live each day to the fullest since there won’t be another one like that.</p><p>We are all running after life’s big pleasures without realizing we are currently walking on life’s priceless treasures.</p><p>Slow down and enjoy the beauty of little things around you. You owe this to yourself.</p><p>Embrace the present moment and cherish it’s precious gift.</p><p><strong>Never sacrifice happiness for achievement</strong></p><p>Goals and dreams are very important for a successful life. Hope for what will appear in the future is what keeps you inspired throughout your journey. Goals are like catalysts that energize our lives. But never sacrifice happiness that is important for well-being to a later time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*QizqhzOOGEejcv-c" /><figcaption>Photo by <a href="https://unsplash.com/@alexblock?utm_source=medium&amp;utm_medium=referral">Alex Block</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Don’t deceive yourself into believing that you’ll be happier once you are wealthy or when you have enough savings to prioritize selfcare and leisure activities. Today is the day to seize the moment and awaken the gift of living. We here in the real world never get a chance to live the life to the fullest.</p><p>Make this moment a new beginning in your life, and focus on what is truely important for you. Do the things that you always wanted to do. Revel in every moment. Listen to your favorite music, explore new language and experiment cooking your favorite dishes. Never put off your happiness for achievements.</p><p>Enjoy your life’s journey and live each day as if it were your last.</p><p><em>“We are all here for some special reason. Stop being a prisoner of your past. Become the architect of your future”-The Monk Who Sold His Ferrari</em></p><h4><strong>Thanks for reading..</strong></h4><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=852d65ff2857" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>