synapse 0.0.1 → 0.2.1

data/client/.Rhistory

@@ -0,0 +1,294 @@
+plot(accepted_denied$x, accepted_denied$y.y / (accepted_denied$y.x + accepted_denied$y.y), type="l", col="blue", xlab="Impressions", ylab="N_denied / N_accepted", main="Denied / accepted ratio")
+library(ggplot2)
+# Search radius (25mi is the default)
+R <- 25.0
+# Blended sort radius, based on bbox
+BLENDED_R <- 16.39
+# Distance is counted in increments of 0.1mi (old code)
+MILES_INCR <- 0.1
+# Take n samples
+RESO_SAMPLES <- 50
+# Or top n by rank
+RANK_TOP_N <- 50
+sf <- subset(read.table("in/sf_points_rank.2012-03-06.csv", sep="\t", header=TRUE), !duplicated(rank))
+resos <- read.table("in/resos_3months_out.2012-03-07.csv", sep=",", header=TRUE)
+sf_resos <- merge(sf, resos, by="hosting_id")
+# Take random samples
+sf_resos <- sf_resos[sample(1:nrow(sf_resos), RESO_SAMPLES, replace=FALSE),]
+# Take top n
+# sf_resos <- subset(sf_resos, rank > 0 & rank <= RANK_TOP_N)
+## Additonal values
+sf_resos$d_R <- sf_resos$distance / R
+sf_resos$o_1 <- sf_resos$nights_1month / 31
+sf_resos$o_2 <- sf_resos$nights_2month / 62
+sf_resos$o_3 <- sf_resos$nights_3month / 93
+for (i in 1:nrow(sf_resos)) {
+# Make log work. < 1 point are all treated as equally bad
+if (sf_resos$points[i] < 1) {
+sf_resos$log_p[i] = 0.0
+}
+else {
+sf_resos$log_p[i] <- log10(sf_resos$points[i])
+}
+}
+# Weights
+W_p <- 0.3 # Note: points aren't normalized. The max is around 2000, so 0.3 * log10() gives just under 1.0
+W_d <- -0.6
+W_o <- -0.2
+sf_resos$weighted_points <- W_p * sf_resos$log_p
+sf_resos$weighted_distance <- W_d * sf_resos$d_R
+sf_resos$weighted_occupancy <- W_o * sf_resos$o_1
+sf_resos$rank2 <- sf_resos$weighted_points + sf_resos$weighted_distance + sf_resos$weighted_occupancy
+# min <- min(sf_resos$rank2)
+# sf_resos$rank2 <- sf_resos$rank2 - min
+MAX_POINTS <- max(sf$points)
+## normalize
+graph_rank <- sf_resos$rank2
+d_R <- data.frame(metric=factor("d_R"), rank=factor(graph_rank), value=sf_resos$distance / R)
+o_1 <- data.frame(metric=factor("o_1"), rank=factor(graph_rank), value=sf_resos$nights_1month / 31)
+o_2 <- data.frame(metric=factor("o_2"), rank=factor(graph_rank), value=sf_resos$nights_2month / 62)
+o_3 <- data.frame(metric=factor("o_3"), rank=factor(graph_rank), value=sf_resos$nights_3month / 93)
+weighted_points <- data.frame(metric=factor("weighted_points"), rank=factor(graph_rank), value=sf_resos$weighted_points)
+weighted_distance <- data.frame(metric=factor("weighted_distance"), rank=factor(graph_rank), value=sf_resos$weighted_distance)
+weighted_occupancy <- data.frame(metric=factor("weighted_occupancy"), rank=factor(graph_rank), value=sf_resos$weighted_occupancy)
+# TODO there should be a better way
+old_points <- data.frame(metric=factor("old_points"), rank=factor(graph_rank), value=sf_resos$points)
+old_blended <- data.frame(metric=factor("old_blended"), rank=factor(graph_rank), value=sf_resos$points)
+old_distance <- data.frame(metric=factor("old_distance"), rank=factor(graph_rank), value=sf_resos$distance * 500)
+log_p <- data.frame(metric=factor("log_p"), rank=factor(graph_rank), value=sf_resos$points)
+for (i in 1:nrow(log_p)) {
+# Make log work. < 1 point are all treated as equally bad
+if (log_p$value[i] < 1) {
+log_p$value[i] = 0.0
+}
+else {
+log_p$value[i] <- log10(log_p$value[i])
+}
+# Old blended
+if (sf_resos$distance[i] < BLENDED_R) {
+old_blended$value[i] = sf_resos$points[i]
+}
+else {
+old_blended$value[i] = log10(sf_resos$points[i] / sf_resos$distance[i])
+}
+}
+rank_data <- rbind(d_R, log_p, o_1, o_2, o_3)
+## end normalize
+old_rank_data <- rbind(old_blended, old_distance)
+# ggplot(data=old_rank_data, aes(x=rank, y=value, fill=metric)) + geom_bar(position=position_dodge())
+new_rank_data <- rbind(weighted_points, weighted_distance, weighted_occupancy)
+filename <- sprintf("out/weighted_blend_components_W_p=%f_W_d=%f_W_o=%f.pdf", W_p, W_d, W_o)
+pdf(filename)
+ggplot(data=new_rank_data, aes(x=rank, y=value, fill=metric)) + geom_bar(position=position_dodge())
+# ggplot(data=rank_data, aes(x=rank, y=value, fill=metric)) + geom_bar()
+library(ggplot2)
+# Search radius (25mi is the default)
+R <- 25.0
+# Blended sort radius, based on bbox
+BLENDED_R <- 16.39
+# Distance is counted in increments of 0.1mi (old code)
+MILES_INCR <- 0.1
+# Take n samples
+RESO_SAMPLES <- 50
+# Or top n by rank
+RANK_TOP_N <- 50
+sf <- subset(read.table("in/sf_points_rank.2012-03-06.csv", sep="\t", header=TRUE), !duplicated(rank))
+resos <- read.table("in/resos_3months_out.2012-03-07.csv", sep=",", header=TRUE)
+sf_resos <- merge(sf, resos, by="hosting_id")
+# Take random samples
+sf_resos <- sf_resos[sample(1:nrow(sf_resos), RESO_SAMPLES, replace=FALSE),]
+# Take top n
+# sf_resos <- subset(sf_resos, rank > 0 & rank <= RANK_TOP_N)
+## Additonal values
+sf_resos$d_R <- sf_resos$distance / R
+sf_resos$o_1 <- sf_resos$nights_1month / 31
+sf_resos$o_2 <- sf_resos$nights_2month / 62
+sf_resos$o_3 <- sf_resos$nights_3month / 93
+for (i in 1:nrow(sf_resos)) {
+# Make log work. < 1 point are all treated as equally bad
+if (sf_resos$points[i] < 1) {
+sf_resos$log_p[i] = 0.0
+}
+else {
+sf_resos$log_p[i] <- log10(sf_resos$points[i])
+}
+}
+# Weights
+W_p <- 0.3 # Note: points aren't normalized. The max is around 2000, so 0.3 * log10() gives just under 1.0
+W_d <- -0.6
+W_o <- -0.2
+sf_resos$weighted_points <- W_p * sf_resos$log_p
+sf_resos$weighted_distance <- W_d * sf_resos$d_R
+sf_resos$weighted_occupancy <- W_o * sf_resos$o_1
+sf_resos$rank2 <- sf_resos$weighted_points + sf_resos$weighted_distance + sf_resos$weighted_occupancy
+# min <- min(sf_resos$rank2)
+# sf_resos$rank2 <- sf_resos$rank2 - min
+MAX_POINTS <- max(sf$points)
+## normalize
+graph_rank <- sf_resos$rank2
+d_R <- data.frame(metric=factor("d_R"), rank=factor(graph_rank), value=sf_resos$distance / R)
+o_1 <- data.frame(metric=factor("o_1"), rank=factor(graph_rank), value=sf_resos$nights_1month / 31)
+o_2 <- data.frame(metric=factor("o_2"), rank=factor(graph_rank), value=sf_resos$nights_2month / 62)
+o_3 <- data.frame(metric=factor("o_3"), rank=factor(graph_rank), value=sf_resos$nights_3month / 93)
+weighted_points <- data.frame(metric=factor("weighted_points"), rank=factor(graph_rank), value=sf_resos$weighted_points)
+weighted_distance <- data.frame(metric=factor("weighted_distance"), rank=factor(graph_rank), value=sf_resos$weighted_distance)
+weighted_occupancy <- data.frame(metric=factor("weighted_occupancy"), rank=factor(graph_rank), value=sf_resos$weighted_occupancy)
+# TODO there should be a better way
+old_points <- data.frame(metric=factor("old_points"), rank=factor(graph_rank), value=sf_resos$points)
+old_blended <- data.frame(metric=factor("old_blended"), rank=factor(graph_rank), value=sf_resos$points)
+old_distance <- data.frame(metric=factor("old_distance"), rank=factor(graph_rank), value=sf_resos$distance * 500)
+log_p <- data.frame(metric=factor("log_p"), rank=factor(graph_rank), value=sf_resos$points)
+for (i in 1:nrow(log_p)) {
+# Make log work. < 1 point are all treated as equally bad
+if (log_p$value[i] < 1) {
+log_p$value[i] = 0.0
+}
+else {
+log_p$value[i] <- log10(log_p$value[i])
+}
+# Old blended
+if (sf_resos$distance[i] < BLENDED_R) {
+old_blended$value[i] = sf_resos$points[i]
+}
+else {
+old_blended$value[i] = log10(sf_resos$points[i] / sf_resos$distance[i])
+}
+}
+rank_data <- rbind(d_R, log_p, o_1, o_2, o_3)
+## end normalize
+old_rank_data <- rbind(old_blended, old_distance)
+# ggplot(data=old_rank_data, aes(x=rank, y=value, fill=metric)) + geom_bar(position=position_dodge())
+new_rank_data <- rbind(weighted_points, weighted_distance, weighted_occupancy)
+filename <- sprintf("out/weighted_blend_components_W_p=%f_W_d=%f_W_o=%f.pdf", W_p, W_d, W_o)
+pdf(filename)
+ggplot(data=new_rank_data, aes(x=rank, y=value, fill=metric)) + geom_bar(position=position_dodge())
+# ggplot(data=rank_data, aes(x=rank, y=value, fill=metric)) + geom_bar()
+install.packages('e1071')
+library("e1071")
+?svm
+data <- read.table("/Users/tobi/Projects/auto_pricing/data.csv")
+data <- read.table("/Users/tobi/Projects/auto_pricing/data.csv", sep=",")
+View(data)
+m <- svm(data$V7, data$V1)
+data <- read.table("/Users/tobi/Projects/auto_pricing/data.csv", sep=",", header=TRUE)
+View(data)
+m <- svm(data$country_state_city_neighborhood, data$price)
+m <- svm(as.factor(data$country_state_city_neighborhood), data$price)
+data_clean <- subset(data, !is.na(country_state_city_neighborhood))
+m <- svm(as.factor(data$square_feet, data$price)
+)
+m <- svm(data$square_feet, data$price)
+x <- seq(0.1, 5, by = 0.05)
+y <- log(x) + rnorm(x, sd = 0.2)
+fix(x)
+m <- svm(x, y)
+new <- predict(m, x)
+plot(x, y)
+points(x, log(x), col = 2)
+points(x, new, col = 4)
+X <- data.frame(a = rnorm(1000), b = rnorm(1000))
+attach(X)
+m <- svm(X)
+View(X)
+m <- svm(~a + b)
+predict(m, t(c(0, 0)))
+data <- read.table("/Users/tobi/Desktop/data.csv")
+data <- read.table("/Users/tobi/Desktop/data.csv", sep="\t")
+View(data)
+data <- read.table("/Users/tobi/Desktop/data.csv", sep=",")
+View(data)
+data <- read.table("/Users/tobi/Desktop/data.csv", sep=",", header=TRUE)
+View(data)
+data$0
+data$1
+data[1]
+data[0]
+data[2]
+plot(data[1], data[2])
+plot(data[2])
+plot(data[1], data[2])
+plot(data[1], data[3])
+library("ggplot2")
+qplot
+qplot(data[1], data[2])
+qplot(x=data[1], y=data[2])
+data <- read.table("/Users/tobi/Desktop/data.csv", sep=",", header=TRUE)
+plot(data)
+data <- read.table("/Users/tobi/Desktop/data.csv", sep=",", header=TRUE)
+library("ggplot2")
+qplot(date, fees, data=data)
+qplot(date, fees, data=data, geom=c("point", "smooth"))
+qplot(date, fees, data=data, geom=c("point", "smooth"), group=1)
+data2011 <- read.table("/Users/tobi/Desktop/data_2011.csv", sep=",", header=TRUE)
+View(data2011)
+qplot(date, fees, data=data2011, geom=c("point", "smooth"), group=1)
+qplot(date, fees, data=data, geom=c("point", "smooth"), group=1)
+qplot(date, fees, data=data2011, geom=c("point", "smooth"), group=1)
+qplot(date, fees, data=data, geom=c("point", "smooth"), group=1)
+data_all <- read.table("/Users/tobi/Desktop/data_all.csv", sep=",", header=TRUE)
+qplot(date, fees, data=data_all, geom=c("point", "smooth"), group=1)
+data_all <- read.table("/Users/tobi/Desktop/data_2011_present.csv", sep=",", header=TRUE)
+qplot(date, fees, data=data_all, geom=c("point", "smooth"), group=1)
+View(data_all)
+data_all <- read.table("/Users/tobi/Desktop/data_all.csv", sep=",", header=TRUE)
+qplot(date, fees, data=data_all, geom=c("point", "smooth"), group=1)
+library("ggplot2")
+data <- read.table("~/all_posts.csv")
+d <- read.tabl('statsd_data.txt')
+d <- read.table('statsd_data.txt')
+View(d)
+r <- signif(d, 2)
+View(r)
+f <- fft(r)
+f <- fft(r$V1)
+fix(f)
+plot(f)
+plot(r)
+describe(r)
+describe(r)
+summary(r)
+stddev(r)
+std(r)
+hist(r)
+hist(r$V1)
+hist(d$V1)
+hist(d$V1, breaks=10)
+hist(d$V1, breaks=30)
+hist(d$V1, breaks=20)
+hist(d$V1, breaks=50)
+hist(d$V1, breaks=60)
+hist(d$V1, breaks=100)
+hist(d$V1, breaks=50)
+summary(d$V1)
+setwd("~/Projects/service_router/client")
+data <- read.table("bench_rewrite_config.dat")
+View(data)
+library("ggplot2")
+qplot(data)
+qplot(data, V3, geom="line")
+qplot(V2, V3, data=data, geom="line")
+qplot(V2, V3, data=data, geom="line", facets=V1)
+qplot(V2, V3, data=data, geom="line", facets=data$V1)
+qplot(V2, V3, data=data, geom="line", facets=as.factor(V1))
+qplot(V2, V3, data=data, geom="line", facets=as.factor(data$V1))
+qplot(V2, V3, data=data, geom="line", facets=as.factor(V1) ~ .)
+qplot(V2, V3, data=data, geom="line", facets=V1 ~ .)
+qplot(V2, V3, data=data, geom="line", colour=V1 ~ .)
+qplot(V2, V3, data=data, geom="line", colour=V1)
+qplot(V2, V3, data=data, geom="line", colour=V1, ylim=c(0, 9999))
+qplot(V2, V3, data=data, geom="line", colour=V1, ylim=c(999, 9999))
+qplot(V2, V3, data=data, geom="line", colour=V1, ylim=c(1000, 6000))
+qplot(V2, V3, data=data, geom="line", colour=V1, ylim=c(1000, 5000))
+qplot(V2, V3, data=data, geom="line", colour=as.factor(V1), ylim=c(1000, 5000))
+qplot(V2, V3, data=data, colour=as.factor(V1), ylim=c(1000, 5000))
+data <- read.table("bench_rewrite_config.dat")
+qplot(V2, V3, data=data, colour=as.factor(V1), ylim=c(1000, 5000))
+qplot(V2, V3, data=data, colour=as.factor(V1))
+qplot(V2, V3, data=data[10:], colour=as.factor(V1))
+qplot(V2, V3, data=data[10], colour=as.factor(V1))
+qplot(V2, V3, data=data[10], colour=as.factor(V1))
+qplot(V2, V3, data=data, colour=as.factor(V1), ylim=c(1000, 25000))
+qplot(V2, V3, data=data, colour=as.factor(V1), ylim=c(1000, 20000))
+data <- read.table("bench_rewrite_config.dat")
+qplot(V2, V3, data=data, colour=as.factor(V1), ylim=c(1000, 20000))
+qplot(V2, V3, data=data, colour=as.factor(V1), ylim=c(1000, 20000))